Bucket.findEntryOffset() can reduce calls to getKeySize()

[swirlds-automation]

🔖 In Bucket.findEntryOffset(), there is no need to compute the key size for the last entry.

...
	// TODO ideally we would only read this once and use in key.equals and here
	// TODO also could avoid doing this for last entry

[swirlds-automation]

_{migrated from:
url=https://github.com/swirlds/swirlds-platform/issues/3932
author:tinker-michaelj, #:3932, createdAt:2021-09-28T15:02:29Z, updatedAt=2023-02-21T23:50:54Z
labels=Performance,Platform Virtual Map,Platform Data Structures,Migration:Data}

[artemananiev]

Skipping call to getKeySize() for the last key in a bucket looks easy, let me do that.

As for getting key sizes and leveraging them in equals(), does it really make sense? I checked all existing virtual key serializers in HS code, some of them use fixed-sized keys (so they just return a constant in deserializeKeySize()), others just read a single byte from the buffer. Not much to save.

OK, if we still decide to calculate key sizes just once, I've got two ideas:

• Call deserializeKeySize() regardless, then pass the size as an extra arg to equalsTo(). Leave it up to implementations to make use (or not) this key size
• Change equalsTo() return value from boolean to int. If it returns 0, the key is considered equal to what's in the buffer. In not 0, it indicates the key size from the buffer. Something like Either<Boolean, Integer> in scala terms

Both ideas don't look very nice to me. Given that we would be able to save just a single call to ByteBuffer.get(), I don't think they are worth implementing. @tinker-michaelj @jasperpotts @OlegMazurov what do you think about it?