Remove fn slot_deltas() from StatusCache

[brooksprumo]

Problem

The slot_deltas() and root_slot_deltas() functions have different behavior w.r.t. not-found slots. They probably shouldn't.

Please see #26170 (comment) for more context.

Summary of Changes

Remove fn slot_deltas()

OLD SUMMARY BELOW

Summary of Changes

Add a new private function that wraps the impl of both slot_deltas() and root_slot_deltas(). Now both have the same behavior (to create a unique entry per missing slot, instead of sharing a single 'empty').

Some benchmarks:

before:

test bench_status_cache_root_slot_deltas  ... bench:       7,712 ns/iter (+/- 388)
test bench_status_cache_slot_deltas       ... bench:       8,826 ns/iter (+/- 1,448)
test bench_status_cache_slot_deltas_empty ... bench:       5,441 ns/iter (+/- 318)

after:

test bench_status_cache_root_slot_deltas  ... bench:       7,596 ns/iter (+/- 344)
test bench_status_cache_slot_deltas       ... bench:       8,582 ns/iter (+/- 539)
test bench_status_cache_slot_deltas_empty ... bench:      27,687 ns/iter (+/- 2,144)

The only change is that 'empty'/missing slots are now more expensive, since a new Status is created for each one. In practice I cannot imagine that roots have zero transactions, therefore this pathological case should never occur.

[ryoqun]

I think the general direction is so good. Discussed elsewhere, i think this pr can further be smaller? could you rework? (sorry for wasted efforts...)

[brooksprumo]

@ryoqun This PR is ready for another review. Thanks!

[ryoqun]

lgtm