Hash-encode keys in StateMap instead of serializing them with a codec

[neysofu]

Description

This PR supersedes #689 and #644. This is the third iteration of the design:

1. Allow Borrow targets as keys as long as they serialize with Borsh. This, as noted, can cause different key types to serialize differently. feat: update StateMap to take borrowee of keys #644
2. Allow all StateKeyEncodePreservingBorrow target types as keys. This was a step in the right direction and maintained the invariants that we're interested in, but it's too easily prone to misuse by SDK users which may inadvertently produce subtle bugs if they're not too careful. Encoding-preserving StateMap borrowee keys #689
3. (This PR) Remove the StateKeyCodec, StateCodec, and StateKeyEncodePreservingBorrow traits, and use Hash to serialize all keys. Borrow targets are always supposed to Hash the same way, and this invariant is relied upon by HashMap of all types so it tends to be respected by all 3rd-party crates. It seems safer than StateKeyEncodePreservingBorrow.

I also made minor changes to StorageKey and CacheKey to allow use as reference, just like #644.

Linked Issues

Related to #689 and #644.

Testing

I'm relying on the current existing tests as no new interfaces were added, only modified.

Docs

I modified the relevant docs for traits and StateMap.

[codecov]

Codecov Report

Merging #718 (335aecd) into nightly (0643f97) will decrease coverage by 0.4%.
The diff coverage is 95.0%.

❗ Current head 335aecd differs from pull request most recent head 102aaef. Consider uploading reports for the commit 102aaef to get more accurate results

Files Changed	Coverage Δ
module-system/sov-modules-api/src/lib.rs	76.9% <ø> (ø)
module-system/sov-state/src/lib.rs	78.2% <ø> (-6.2%)	⬇️
module-system/sov-state/src/internal_cache.rs	64.7% <78.5%> (ø)
module-system/sov-state/src/storage.rs	94.9% <93.0%> (-4.1%)	⬇️
module-system/sov-state/src/scratchpad.rs	94.0% <96.9%> (+<0.1%)	⬆️
module-system/sov-state/src/map.rs	98.5% <97.7%> (+5.7%)	⬆️
...-system/module-implementations/sov-evm/src/call.rs	94.5% <100.0%> (+<0.1%)	⬆️
...system/module-implementations/sov-evm/src/query.rs	55.0% <100.0%> (ø)
...le-system/sov-modules-api/src/default_signature.rs	60.2% <100.0%> (+0.8%)	⬆️
module-system/sov-state/src/codec.rs	72.2% <100.0%> (+20.5%)	⬆️
... and 3 more

... and 3 files with indirect coverage changes

[vlopes11]

The approach looks good! I have one question on why we still use serialization, even after the introduction of the Hasher of stdlib

[vlopes11]

Very good, will help a lot with the ergonomics