[FEATURE] High-resolution fractional bucketing in flagd provider

[toddbaert]

Sub-issue of open-feature/flagd#1903. See the parent issue and ADR for full details.

Replace the current 100-bucket float-based fractional bucketing with integer arithmetic using MaxInt32 (2,147,483,647) as the max weight sum:

bucket = (uint64(hashValue) * uint64(totalWeight)) >> 32

Weights must be integers and sum to at most MaxInt32. Java: convert hash to unsigned via hash & 0xFFFFFFFFL, use long for multiply, >>> 32 (unsigned shift).

This is a behavioral breaking change (bucket assignments will shift), but the API surface is unchanged.

⚠️ Dependency: complete the nested fractional evaluation feature first: #1737

@toddbaert will coordinate the release across providers.

[toddbaert]

@chrfwow if you want to take this, I suggest you do #1737 at the same time, or first.

You will have some bad git conflcits with somebody else, otherwise.