Fix try_shrink not freeing back to pool

[cetra3]

Which issue does this PR close?

No issue raised

Rationale for this change

try_shrink never released any memory from the pool, and so would cause a "memory leak"

What changes are included in this PR?

A few changes to MemoryReservation

Are these changes tested?

Yes & new test has been added

Are there any user-facing changes?

Nope

[comphead]

@andygrove FYI

[comphead]

Thanks @cetra3 that looks like a really nice catch, I checked the test currently failing on main for other mem pools as well.

[hareshkh]

@cetra3 @mbutrovich @comphead : Any chance we could get this in a patch to branch-51? I can run the backports but would be interested to have this fix there. (I can backport to branch-52 as well)

[comphead]

@cetra3 @mbutrovich @comphead : Any chance we could get this in a patch to branch-51? I can run the backports but would be interested to have this fix there. (I can backport to branch-52 as well)

We should prob wait for DF 53 release which should be pretty soon
#19692

[mbutrovich]

@cetra3 @mbutrovich @comphead : Any chance we could get this in a patch to branch-51? I can run the backports but would be interested to have this fix there. (I can backport to branch-52 as well)

I would be interested in a backport to DF52. I will bring it here: #20287. If you open the backport PR for 52, could you add it to that tracking issue? I am not sure if we are planning another 51 patch release.

[hareshkh]

@mbutrovich @comphead : Sorry, my bad here. This PR only fixes a bug introduced by #19759 so branch-51 and branch-52 (which use usize instead of AtomicUsize) do not have the bug.
In branch-52, try_shrink:

pub fn try_shrink(&mut self, capacity: usize) -> Result<usize> {
    if let Some(new_size) = self.size.checked_sub(capacity) {
        self.registration.pool.shrink(self, capacity);
        self.size = new_size;
        Ok(new_size)
    } else {
        internal_err!(
            "Cannot free the capacity {capacity} out of allocated size {}",
            self.size
        )
    }
}

Which already calls pool.shrink() correctly.

I was debugging another issue I am running into and I thought this would be the fix for branch-51/branch-52 as well but that is not the case :(