Compile failure with PCH on MSVC for ARM

[Flamefire]

I see downstream failures on CI using the GHA windows-11-arm runner: https://github.com/boostorg/random/actions/runs/34051005391/job/101534518823?pr=157

2026-09-06T18:24:34.2418803Z assoc_laguerre.cpp
2026-09-06T18:24:34.2421303Z c1xx: error C3859: Failed to create virtual memory for PCH
2026-09-06T18:24:34.2425018Z c1xx: note: PCH: Unable to get the requested block of memory
2026-09-06T18:24:34.2452229Z c1xx: note: consider using /Fp to allow the compiler to reserve the memory early
2026-09-06T18:24:34.2462618Z c1xx: note: please visit https://aka.ms/pch-help for more details
2026-09-06T18:24:34.2507344Z c1xx: fatal error C1076: compiler limit: internal heap limit reached

Intel is also failing: https://github.com/boostorg/random/actions/runs/34051005391/job/101534519645?pr=157
But that seems like a B2 issue

Maybe update the GHA CI to use the Boost.CI reusable workflow to get more configurations tested.

[mborland]

Can Boost.CI handle running individual test suites? Right now we have the 14 of them broken down across CI runs due to resource usage? Agree the Intel one looks like a B2 problem, and the windows-11-arm one looks like OOM.

[Flamefire]

The reusable one?
You can set b2_targets to e.g. libs/math/test//github_ci_block_1 to run a specific target
There is also include_jobs to configure specific job configs to use.
And timeout if the job still takes too long.