Remove FlexZeroVec

[Manishearth]

We don't really use it anymore I think, except potentially for some usize ZHMs, which we can instead use Index8 VZV backing stores for if we desire.

[Manishearth]

@sffc one lingering use case is in the ZeroTrie benches:

icu4x/utils/zerotrie/benches/overview.rs

Line 101 in 7634c7e

let actual = actual.map(|x| <zerovec::vecs::FlexZeroSlice as zerovec::maps::ZeroVecLike<usize>>::zvl_get_as_t(x, |y| *y));

This is the usize case. I'm happy with getting rid of it for now.

I don't really think it makes sense for usize to special-case as a flex container anyway.

[Manishearth]

The other breakage is in BlobSchemaV1, which is probably going to go away in #5603

icu4x/provider/blob/src/blob_schema.rs

Line 80 in 7634c7e

pub markers: ZeroMap2dBorrowed<'data, DataMarkerPathHash, Index32U8, usize>,

[Manishearth]

This can't be landed until we fix #5603

[Manishearth]

The bulk of the FZV usage is in the ZeroTrie benches 🙃

[sffc]

Suggested change

	g.bench_function("ZeroMap/usize", |b| {
	g.bench_function("ZeroMap/u32", |b| {

[sffc]

Suggested change

	g.bench_function("ZeroMap/usize", |b| {
	g.bench_function("ZeroMap/u32", |b| {

[sffc]

What changed across the force-push? I see a bunch of changes pulled in from main.

[Manishearth]

@sffc rebase over main and fix the zerotrie builder test numbers broken by the VZV changes

[sffc]

I think the two diffs you applied earlier got undone

[Manishearth]

Oops, fixed. Forgot I applied those patches late at night