GUI: Add "Generate debug log" action - #3328

Introduce a reusable debug text dialog and helper functions for generating and displaying DeepLabCut diagnostic reports. Adds deeplabcut/gui/dialogs/debug_dialog.py implementing DebugTextDialog (read-only log view with refresh/copy/keyboard shortcut), providers to render logs and full issue reports, show_debug_report_dialog to install/reuse recorders and present the report, and create_generate_debug_log_action to wire up a QAction. Also exports these symbols via deeplabcut/gui/dialogs/__init__.py. Dialog instances are attached to the parent to avoid GC and duplicate windows.

Import and install a debug recorder and add a "Generate Debug Log" action to the Help menu. The patch imports create_generate_debug_log_action and install_debug_recorder, creates a recorder (logger_name="deeplabcut"), builds the debug-log action with specified options (include_module_paths=False, include_executable_paths=True, log_limit=1000), inserts it into the Help menu with a separator, and adjusts menu sizing. This allows users to generate debug logs from the GUI.

Add a new test module for deeplabcut.core.debug covering InMemoryDebugRecorder and log_timing. Tests verify install_debug_recorder idempotency and retrieval, message and exception capture, bounded buffer behavior, render_text contents, clear() resetting records and drop count, and log_timing behavior when enabled, disabled, and with thresholding. Uses pytest fixtures to create isolated logger namespaces and monkeypatch to simulate timings and logger behavior.

Add project/license header comments to deeplabcut/core/debug/__init__.py and deeplabcut/gui/dialogs/__init__.py. In core/debug/__init__.py, clean up the public API exports: remove internal symbols DLC_LOG_TIMING and LOG_QUEUE_MAXLEN from __all__, reorder entries, and ensure collect_executable_summary is exported. These changes clarify public exports and add attribution/licensing headers.

Replace generic "GUI" and "console" loggers with namespaced loggers ("deeplabcut.gui" and "deeplabcut.gui.console"). Store a console logger on the MainWindow instance and use it for setting level and printing status messages. This standardizes logger names and avoids relying on a global "console" logger.

Add DebugTextDialog.update_content to update window title, text provider, and hint when reusing an existing dialog instance. Update _get_or_create_debug_dialog to call update_content for existing dialogs instead of recreating them. Rename the initial_hint parameter to text_hint and adjust the show_debug_report_dialog call site accordingly. This lets the debug dialog refresh its metadata without reconstructing the widget.

Add environment-driven debug/timing configuration and make the in-memory debug recorder less intrusive. Changes:
- deeplabcut/core/debug/_debug_utils.py: add license header, import os, add reload_debug_settings_from_env(), and helper parsers _env_flag and _env_optional_float to read boolean/optional-float env vars.
- deeplabcut/core/debug/debug_logger.py: add license header, read DLC_LOG_TIMING and DLC_LOG_TIMING_THRESHOLD_MS from environment, import the new env helpers, and use an effective_threshold in log_timing().
- install_debug_recorder(): add handler_level and ensure_logger_level params, use handler_level for the recorder, and only lower the logger level when explicitly requested.
- Adjust external tools section reporting logic to avoid misleading reporting when executables is None.

Rationale: allow debug timing and thresholds to be controlled via environment variables, make recorder installation configurable without forcing global logger verbosity, and improve robustness of external-tool reporting.

Move the reload_debug_settings_from_env function from deeplabcut/core/debug/_debug_utils.py into deeplabcut/core/debug/debug_logger.py. The function (which reloads DLC_LOG_TIMING and DLC_LOG_TIMING_THRESHOLD_MS from environment variables) is removed from _debug_utils.py and added to debug_logger.py to colocate timing/debug env handling with the logging code. No behavioral changes intended.

Import QApplication and set an override WaitCursor in DebugTextDialog.refresh_text while calling the text provider, restoring the cursor in a finally block. This provides visual feedback during potentially long debug-text generation and ensures the cursor is always reset even if the provider raises an exception.

Allow the in-memory debug recorder to optionally auto-initialize the target logger's level. Import typing.Literal and change install_debug_recorder's ensure_logger_level parameter to accept None, int, or the new "auto" sentinel (default). Expand the docstring to document behavior: "auto" sets the logger to handler_level only when the logger has no explicit level (NOTSET); int lowers the logger level only if more restrictive; None leaves the logger untouched. Adjust the logic accordingly and update the GUI caller to pass handler_level and ensure_logger_level="auto" when installing the recorder.

Convert libraries and executables to tuples in make_issue_report_provider so the returned provider can be called repeatedly even if given a one-shot iterable. Comment out redundant initial refresh_text() calls in the dialog (showEvent triggers the initial refresh). Also set logger.propagate = False in the test cleanup to avoid log propagation during tests.

Change install_debug_recorder to remove the special "auto" behavior for ensure_logger_level and make its default None; now only an int will lower the logger level when explicitly requested. Simplify some docstrings and clarify capacity wording. Add LibrarySpec.prefer_module_version and a new _module_version helper to prefer a module's __version__ when collecting versions (falling back to the distribution version if unavailable). Mark opencv-python to prefer the module version and update collect_version_summary to use the resolved module name for module paths.

Expose DebugSection from deeplabcut.core.debug by adding it to the module imports and __all__. Expand test_debug_logger.py with extensive tests for building and formatting debug reports, version and executable summaries, and log handling (using monkeypatched helpers and InMemoryDebugRecorder) to validate report contents and edge cases.