[bugfix] re-aligning subjfloat_functions with standardFunctionInfos when BFLOAT16 undefined

[nibrunieAtSi5]

Work in progress

standardFunctionInfos uses #ifdef BFLOAT16 to guard the addition of 2 entries from bf16 to/from f32 conversions, while subjfloat_functions array did not, this could lead to a misalignment depending on the build configuration.

This could have been introduced in https://github.com/ucb-bar/berkeley-testfloat-3/commit/b205774f3c3358d531cd8aa613914da101e17534, to fix another issue (or maybe I am misinterpreting the issue)

IMHO, It would be good to switch to a single array structure to avoid the possibility of creating discrepancies between the two (and also having to manually enforce ordering).