#2974: Add hash-based journal selection for improved ledger distribution

[jprieto-temporal]

Main Issue: #2974

Motivation

When multiple journals are configured, BookKeeper selects a journal using ledgerId % numJournals. This works well when
ledger IDs are random or sequential, but produces poor distribution when ledger IDs follow strided patterns (e.g., when
multiple clients allocate IDs from disjoint ranges). Uneven journal utilization can cause hotspots and reduce write
throughput.

Changes

• Add journalHashBasedSelection configuration option (default: false)
• When enabled, use Fibonacci hashing on ledger IDs before journal selection
• The hash function multiplies by the golden ratio constant and folds high bits into low bits to break predictable
  patterns
• Add integration test verifying read/write correctness with hash-based selection

[eolivelli]

what about putting this block into a static method so that we can add unit tests ?

The function gets the index and the number of journals and computes the journal index

[jprieto-temporal]

Thank you for your review. I added the tests and factored out the static method.

[dlg99]

there are other places where ledgerid is used for routing (e.g. ordered thread executor).
Related change: #3023
Maybe it is worth refactoring the selection? Also the PR has a test you can try to reuse

[jprieto-temporal]

@dlg99 not opposed to factoring it out and using the same hashing that is used in this PR since it provides better distribution in the general case, but I think that could be done in a separate PR.