core: (Assembly Format) fix parsing for keyword/punctuation (#6425)

Fixes the `parse` and adds `parse_optional` methods for
`KeywordDirective` and `PunctuationDirective`, making these no longer
accept the empty string outside of optional groups.

This revealed two bugs in `arm_neon` and `tosa`. I've tried to fix both
in the a way that seems to make the most sense to me, but it would be
good if these could be checked.

Stacked on top of #6424

dialects: (stim) Remove stim (#6375)

core: Add float-semantics to support low-precision (#6254)

Extends support for reduced-precision floating point values by adding
support for reduced-precision literals, and a central mechanism to
implement a variety of different flavours. The types itself were added
in #6250. This PR adds support for literals, by introducing the
`FloatSemantics` class based on MLIR's `fltSemantics` concept (alongside
`fltNonfiniteBehavior` and `fltNanEncoding`).

Stack:
* #6250 
* #6256 
* #6263

testing: add test for hello world LLVM IR emission (#6241)

CI: enable ruff RUF015 rule (#6223)

dependencies: update pre-commit hook astral-sh/ruff-pre-commit to v0.…

…15.16 (#6158)

backend: (llvm) add vector.reduce.fadd and fmul backend conversion (#…

…6120)

Add llvmlite backend conversion for llvm.intr.vector.reduce.fadd and
llvm.intr.vector.reduce.fmul.

Depends on #6131.

Refs:
- LLVM LangRef fadd:
https://llvm.org/docs/LangRef.html#llvm-vector-reduce-fadd-intrinsic
- LLVM LangRef fmul:
https://llvm.org/docs/LangRef.html#llvm-vector-reduce-fmul-intrinsic

core: (Constraints) ParamAttrConstraint to BaseAttr (#6078)

Convert a `ParamAttrConstraint` with only `AnyAttr` constraints for the
parameters to a `BaseAttr` Constraint. This allows a bunch of `constr`
methods to be simplified, as we were doing the same check in each one.

dialects: (arith) Added bf16 as floating-point-like (#5918)

Currently floatingPointLike in the arith dialect is limited to FP16,
FP32 and FP64. However, in MLIR this uses the AnyFloatConstr and indeed
the xDSL math dialect does the same too. Have therefore moved to this
same approach, which means that bf16, fp80 and fp128 are also supported
by arith conversion and floating point operations which is inline with
MLIR. Have added filecheck tests for these.

dialects: (riscv_snitch) add vfsubh and vfsubs (#5821)