bgc deadlock fix

[Maoni0]

this was a regression introduced in #105545. this would only manifest during the beginning of a process when we needed to grow the # of heaps (and can grow) and we haven't done a gen2 GC yet so we set it to doing one.

the problem is the check I made did not include the ephemeral GC that may happen at the beginning of a BGC before we set gc_background_running to true. so at the end of that eph GC, we are in calculate_new_heap_count and set trigger_initial_gen2_p to true, not realizing we are already in a BGC.

then during the joined_generation_to_condemn at the beginning of the next GC, if our conclusion was still doing an eph GC we'd make it a BGC due to trigger_initial_gen2_p which obviously would cause problem if a BGC is already in progress (I do have an assert for this but we haven't seen this in a dbg build...).

the fix is to simply use the right check - is_bgc_in_progress() instead of background_running_p() which includes that eph GC case.

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/gc
See info in area-owners.md if you want to be subscribed.

[Maoni0]

/backport to release/9.0

[github-actions]

Started backporting to release/9.0: https://github.com/dotnet/runtime/actions/runs/11287169324