Searching leader schedule for leader slots is inefficient

[carllin]

Problem

Follow up to: #8468

In the current implementation here: https://github.com/solana-labs/solana/pull/8468/files#diff-bf2148039ffb77b38df2bbae2f74b418R146,

Once the first leader slot N is found, the last slot is at most N + GRACE_TICKS_FACTOR * MAX_GRACE_SLOTS. This means:

1. We iterate through at most N + GRACE_TICKS_FACTOR * MAX_GRACE_SLOTS
2. We only do at most GRACE_TICKS_FACTOR * MAX_GRACE_SLOTS lookups in RocksDb because a validator only does a lookup if the slot belongs to them.

What still sucks: You still have to iterate through N entries in the schedule, granted these iterations are a lot faster because they do not incur a RocksDb lookup.

Proposed Solution

Build a BTreeSet with all the leader schedule slots a validator is scheduled for, which should make "next" queries fast with the btreeset.range function. Purge the set when roots are set.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs.

[carllin]

Resolved by #15249