feat: OpenMP + AVX-512 bindings and calcStructuresParallel() for trajectory analysis

[aminkvh]

Summary

Adds OpenMP and SIMD compile flags to setup.py, and exposes the new
freesasa_calc_structures_parallel() C API as calcStructuresParallel()
in Python. This enables near-linear speedup for MD trajectory analysis.
All existing tests pass.

What Changed

• setup.py — Auto-detects and enables -fopenmp, -march=native,
  -O3, -ffast-math at build time. Graceful fallback if unavailable.
• src/freesasa.pyx — New calcStructuresParallel(structures, params)
  function. Takes a list of Structure objects, returns a list of Results.
• src/cfreesasa.pxd — C declaration for freesasa_calc_structures_parallel
• lib/src/ — Embedded parallel C sources (replaces submodule reference)
  from the companion fork: https://github.com/aminkvh/freesasa-parallel

Usage

params = freesasa.Parameters()
params.setNThreads(8)   # 8 frames processed simultaneously

results = freesasa.calcStructuresParallel(frame_structures, params)
total_areas = [r.totalArea() for r in results]

Performance (8 frames × 58,674 atoms)

Mode	Time	Speedup
Serial (original)	9.38 s	1.0×
Parallel 4 frames	2.52 s	3.7×
Parallel 8 frames	1.38 s	6.8×

Tests

24/24 compatibility tests pass covering the full existing Python API.