CDRIVER-5641: Build fix for _FORTIFY_SOURCE

[mdbmes]

This is a follow-up for #1868 to fix a build break in the abi-compliance-check task.

The break isn't straightforward to reproduce. It seems to require gcc-9(?), -D_FORTIFY_SOURCE=2 and -Og.

The cause is a set of tests that intentionally generate an oversized memcpy within dead code, to ensure the surrounding guard prevents the binary vector read/write operation. When _FORTIFY_SOURCE is set, additional checks are enabled for string functions like memcpy, enforcing a maximum object size of SIZE_MAX/2. In this patch, any time we build with _FORTIFY_SOURCE the maximum counts for these memcpy are adjusted so we don't overflow even for the largest element type (float32).

[mdbmes]

Here's the abi-compliance-check evergreen task, it's not included in the usual builds for pull requests:
https://spruce.mongodb.com/version/67ca1d3445a9cf0007b19064/tasks

[eramongodb]

Interesting find. It looks like the details involve use of Object Size Checking Built-in Functions (e.g. for memcpy) and selecting "a closest surrounding subobject" as "the object a pointer points to" when _FORTIFY_SOURCE > 1. However, I am unable to determine why this behavior may be restricted to specific GCC versions.

[mdbmes]

why this behavior may be restricted to specific GCC versions

I'm guessing both the gcc version and the -Og dependency relate to dead code elimination maybe?
Thanks for the review!