Fix panic in ProgressMap when the underlying block may have been dumped and repaired

[wen-coding]

Problem

During today's testnet accident, we saw validators panicking in ProgressMap when underlying blocks have been purged:

[2024-08-20T07:07:42.166017158Z WARN solana_core::replay_stage] purging slot 288668257
[2024-08-20T07:07:43.020116943Z WARN solana_core::replay_stage] purging duplicate descendant: 288668259
thread 'solReplayStage' panicked at core/src/consensus/progress_map.rs:294:32:
slot=288668259 must exist in ProgressMap
0: rust_begin_unwind
1: core::panicking::panic_fmt
2: solana_core::consensus::progress_map::ProgressMap::get_propagated_stats_must_exist
3: solana_core::consensus::progress_map::ProgressMap::get_leader_propagation_slot_must_exist
4: solana_core::replay_stage::ReplayStage::retransmit_latest_unpropagated_leader_slot

Proposed Solution

Audit all _must_exist function and fix.