JIT compile regression

[alessandrod]

As part of working on #3481 we discovered that program_cache_us times seem to have regressed in 2.0.

#3481 (comment)

Something is off in the executor metrics too, jit_compile_us doesn't seem to be aggregated correctly.

[alessandrod]

@bw-solana @Lichtso @buffalojoec moving to this issue since I landed and closed the original PR

@Lichtso this smells like a JIT regression in rbpf

Also someone fix the jit metrics pls

[buffalojoec]

@alessandrod I'll look into metrics asap

[Lichtso]

Um, are we sure that it is in v2.0? Because so far I have only seen comments about v2.1

[alessandrod]

Left is master, right is 2.0, so yeah seems to be in 2.0 too

[alessandrod]

after a few hours looks like it got significantly worse