gh-131798: JIT inline function pointer address for extended binary op

[kumaraditya303]

• Issue: Better uop coverage in the JIT optimizer #131798

[Fidget-Spinner]

Do you know this is worth it? In this case it's only one dependent load, and something that always resolves the same. So I'm not too sure. We try not to add new uops into the optimizer if they're not worth it as it's more maintenance burden sometimes.

[eendebakpt]

I benchmarked the optimization in this PR on top of #128956

Run 1

Benchmark	bench_base	bench_turbo
tuple_concat_unpack	92.5 us	92.9 us: 1.00x slower
str_repeat	99.2 us	97.8 us: 1.01x faster
bytes_concat	102 us	103 us: 1.01x slower
bytes_repeat	89.0 us	87.0 us: 1.02x faster
tuple_repeat	85.2 us	83.8 us: 1.02x faster
chained_list_ops	143 us	140 us: 1.02x faster
mixed_float_int	49.6 us	50.8 us: 1.02x slower
float_mix_mul	50.9 us	51.4 us: 1.01x slower
Geometric mean	(ref)	1.00x faster

Not significant: list_concat_subscr, dict_merge.

Run 2

Benchmark	bench_base2	bench_turbo2
list_concat_subscr	95.5 us	94.7 us: 1.01x faster
tuple_concat_unpack	92.5 us	93.0 us: 1.01x slower
str_repeat	99.0 us	98.1 us: 1.01x faster
bytes_concat	102 us	103 us: 1.01x slower
bytes_repeat	89.3 us	87.2 us: 1.02x faster
tuple_repeat	85.3 us	83.7 us: 1.02x faster
dict_merge	181 us	182 us: 1.01x slower
chained_list_ops	142 us	140 us: 1.02x faster
mixed_float_int	49.6 us	50.8 us: 1.02x slower
Geometric mean	(ref)	1.00x faster

Numbers are close to (or in) the noise. Maybe 1% faster for this kind of ops.

[github-actions]

This PR is stale because it has been open for 30 days with no activity.