DOC: Add JOSS paper draft - #968

- Mark Xarray/Zarr integrations as optional
- Note the code example is condensed (full version in docs)
- Tie coverage/typing claims to CI and Codecov
- Source the download statistics
- Note open-source adopters are independent of the author
- Describe SATP citation as a methodology repository

ConTeXt (the JOSS PDF engine) mis-renders the graphviz SVG, causing
overlapping text. Switch to a vector PDF figure, which embeds cleanly.

Add the on-demand single-output execution mode (calling the pipeline
directly, optionally supplying intermediate results) alongside the
parallel map-reduce mode.

Add a partial reduction (charge[:, j] -> avg_over_V_left[j]) alongside
the full scalar reduction, demonstrating that mapspec reduces over a
chosen axis while keeping others. Regenerate the figure accordingly.

Replace the placeholder charge/reduction with a real linear-electrostatics
model: induced charge Q = C_left*V_left + C_right*V_right, with the
capacitances derived from geometry and dielectric. The partial reduction
now fits dQ/dV_left to recover the mutual capacitance (2D->1D), and the
full reduction reports the peak induced charge. Verified the exact snippet
runs and recovers the analytic capacitance; regenerate the figure.

Collapse the capacitance computation to one line, drop redundant inline
comments now covered by the prose, and remove verbose annotations. Same
behavior and figure; ~8 fewer lines.

- Correct overhead to 10 us with hardware/benchmark context (matches docs)
- Lead Research impact with independent adoption; downloads now support it
- Merge a redundant framework-agnostic sentence in Statement of need
- Polish the Zarr citation

State the GitHub/Discord facts directly rather than asserting an active
community, which is harder to defend.

- Fix broken affiliations block: index 2 was referenced but undefined.
  Restore TU Delft (affiliation) + IonQ (current address) per intent.
- Name the Adaptive Scheduler backend for SLURM submission.

Single affiliation (IonQ), index 1; removes the dangling index-2
reference.

The FAQ this line links to states ~10 us on an M4; update the README to
match and stay consistent with the JOSS paper.

The SATP-Methodology repo is a reproducible research manuscript (Mitchell
& Aletta, soundscape circumplex translations) whose data analysis runs on
pipefunc. Update the citation and prose to reflect that it is a real study
producing results with pipefunc, not just a methodology repo.

Add the Applied Acoustics paper (doi 10.1016/j.apacoust.2024.110109) for
the study and keep the reproducible-analysis repo as a separate reference,
since the pipefunc-based repo postdates the paper (to be confirmed with the
authors).