feat: HashBuilder::add_leaf_unchecked

[hai-rise]

We're stress-testing live state root calculation in Reth that led back to this crate (thanks for the great work!).

So, this assert takes ~7.7% of HashBuilder::add_leaf and ~1% of StateRoot::calculate:

trie/src/hash_builder/mod.rs

Lines 113 to 114 in 25c252f

	pub fn add_leaf(&mut self, key: Nibbles, value: &[u8]) {
	assert!(key > self.key, "add_leaf key {:?} self.key {:?}", key, self.key);

This PR adds an unchecked version for performance-critical users who guarantee keys are added in sorted order (like Reth with its node iterator).

[mattsse]

alternatively we could convert this to a debug_assert

thoughts @DaniPopes @rkrasiuk ?

[DaniPopes]

We can have this, but we should have the assert be a debug_assert in the unchecked function too

[hai-rise]

I've added the debug_assert for now. Should I look into CI and should we add HashBuilderError like ProofVerificationError in the future?

[rkrasiuk]

supportive of keeping assert in reg method and having unchecked with debug_assert