…tests for syntax validation

- Added PNG artifacts for effect preservation tests.
- Implemented unit tests for OBSLogManager to verify script status and completion detection.
- Created unit tests for scenario loader to ensure scenarios load correctly and handle errors appropriately.

- Make find_package(PNG) optional so Windows can resolve it transitively
  through OBS::libobs, matching the behavior before this PR was raised
- Fix absolute module path in test_network_simulation.py: use
  util.generate_packets instead of tests.e2e.util.generate_packets
  since pytest runs from tests/e2e/

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

The tests/script build also had find_package(PNG REQUIRED) which
fails on Windows CI. Made it optional to match the top-level fix.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

S_ISREG is a POSIX macro not available on MSVC. Add a compatibility
definition using the Windows _S_IFMT/_S_IFREG constants so the file
stability check compiles on Windows.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

The ASSERT IMAGE_EQUALS feature uses libpng, but on Windows the
obs-deps PNG .lib is not found by CMake's FindPNG module (headers
are present but PNG_LIBRARY is missing). This caused 25 unresolved
external symbol errors when linking c64stream.dll.

Fix by:
- Defining compile-time symbol C64_HAVE_PNG when PNG::PNG is linked
- Guarding png.h include, all PNG helper structs/functions, and the
  OP_ASSERT_IMAGE_EQUALS case body with #ifdef C64_HAVE_PNG
- Returning "PNG support not available" on platforms without PNG

…e resolution

- Implemented PreserveSizeCanvasMatchAssertion to validate OBS canvas composition against preserve-size semantics.
- Added new assertion to the assertion runner and included it in the ntsc_preserve_compare scenario.
- Created unit tests for the PreserveSizeCanvasMatch assertion covering frame index computation, source screenshot dimension validation, and canvas geometry checks.
- Introduced utility functions for generating test PNGs and synthetic canvas frames.
- Ensured that tests can detect regressions related to preserve-size handling in OBS.

The Python Unit Tests CI job was failing because:
- validate_source_dimensions() uses Pillow (PIL) to read PNG dimensions
- Pillow was not listed in tests/e2e/requirements.txt
- The function correctly returns status='skip' when PIL is absent, but
  the unit tests asserted on ok/message without checking for skip first

Fix:
- Add Pillow>=10.0.0 to tests/e2e/requirements.txt
- Guard TestSourceDimensionValidation tests with skipTest on skip status

…ed tests

Add three new validation checks to quantitative scanline mode:

Phase 5b - Gap group count: when expected_gap_groups is specified in
thresholds, validate the observed number of gap groups matches (±tolerance).
For NTSC 240 pixel rows, expect exactly 239 gaps (one between each
consecutive pair of C64 pixel rows, CRT electron-beam model).

Phase 5c - Gap width uniformity: all gap bands must have the same pixel
width (±1 px for OBS scaling rounding). Catches rendering artifacts that
produce irregular gap sizes.

Frame extraction: score by gap group count instead of BLACK row count.
This naturally selects frames rendered at virtual resolution
(preserve_size=false) which have the full one-gap-per-C64-pixel-row
topology (239 groups), over logical-resolution frames (80 groups).

Applied to ntsc_preserve_compare scenario with expected_gap_groups: 239.

…aling

…s for CRT effects