don't use ffiresult in monomorphize functions

[Shoeboxam]

Through the process of writing this, I've become more unsure about switching these. On one hand, it's nice to be able to try, but on the other hand,

• the error handling is now inconsistent with the root code
• the macro-based error handling is easier to use with proc-macro-generated dispatch, should we change our dispatch to it

Alternatively, it may not matter-- as the proc macro could simply enclose the dispatch in a closure.

[Shoeboxam]

• don't use ffiresult in monomorphize functions #921 👈
• tweak FFI traits for more helpful compile errors #917 : 1 other dependent PR (#763 )
• fallible MetricSpace check #916
• user-friendly downcast-ref error message #907
• support multithreading #874
• make_base_discrete_exponential proof #705
• add make_base_discrete_exponential ffi #754
• add make_base_discrete_exponential #704
• make_quantile_score_candidates proof #703
• add make_quantile_score_candidates ffi #753
• add make_quantile_score_candidates #702
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

Join @Shoeboxam and the rest of your teammates on Graphite

[andrewvyrros]

It's a little tricky, but I think it's still a solid overall improvement. Moving to proc macros is still an option in the future to resolve remaining issues.

The base branch was changed.