[zfergus]

Description

Builds the Windows Debug CI job with /O2 /Ob1 instead of /Od /Ob0 /RTC1, following the Linux/macOS -O1 change in #301. Only .github/workflows/continuous.yml changes.

• After Fix stale ccache keys and slow Debug CI tests #301, Windows Debug is the longest CI job: its Tests step normally takes 1,100–2,000 s.
• Summed over all tests, Windows Debug spends 6,258 s vs 468 s for Windows Release (13×). Linux Debug went from 5,311 s at -O0 to 171 s at -O1.
• There are two separate costs:
  • No inlining at /Od: Eigen-heavy tests suffer as they did on Linux -O0, e.g. GCP rest-shape measure at 1,114 s.
  • The Debug runtime's STL iterator checks: Point-triangle distance type takes 403 s vs 24 s on Linux -O0, and LBVH::build takes 435 s vs 60 s.
• /O2 removes the first cost. /Ob1 still inlines Eigen's small functions but avoids the extra compile time of /Ob2.

Impact:

• Kept: asserts (NDEBUG stays undefined) and the STL iterator checks (Debug runtime), which are what Windows Debug catches that Linux Debug doesn't.
• Lost: /RTC1's stack-frame and uninitialized-variable checks, since MSVC can't combine them with optimization. The ODR check job still covers the Eigen ABI stack-smashing case.
• Cold builds get slower: somewhere between today's Debug (~750 s) and Release (~1,800 s). Warm ccache builds from main (enabled by Fix stale ccache keys and slow Debug CI tests #301) absorb this after the first push.

Type of change

• Enhancement (non-breaking change which improves existing functionality)

How Has This Been Tested?

• Per-test timings taken from the CI logs of Fix stale ccache keys and slow Debug CI tests #301 (Windows Debug, Windows Release, Linux Debug -O0 and -O1).
• Windows Debug Tests-step times across the last 19 runs range from 692 s to 1,983 s; build time scales with them, so the variance comes from the runners.
• continuous.yml parses as YAML.
• The speedup itself is only measurable on this PR's CI (no local Windows machine). This PR's build will be cold, but the Tests step isn't affected by caching.

Checklist

• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• My changes generate no new warnings
• New and existing unit tests pass locally with my changes