feat(rust)!: removed lossy type coercion for primitive numeric types …

[dhvish]

…(#25624)

From what I understand, is_in currently falls back to lossy supertype conversion with try_get_supertype if the lossless upcast ( get_numeric_upcast_supertype_lossless ) doesn't work.

I made these assumptions in my changes to resolve_is_in

• return an error if no lossless supertype exists for the two primitive numerics.
• kept the existing Null behavior with try_get_supertype.
• removed the a != b branch since the a == b match arm ensures that a != b will always be true (right?).

Let me know if this aligns with the intended direction. I'll fix stale/add tests for this once I get the confirmation.

Currently these tests fail. Most of these make sense as they involve lossy numeric casting.

[orlp]

The operation should still be supported, but be done properly. The lack of a good supertype is no excuse.