stagedsync: add aggressive prune timeout for forkchoice finality mode

When --experimental.use-forkchoice-finality is enabled, the node stays
stable at tip and no longer falls back to syncToTip mode, which
previously triggered initialCycle=true with aggressive pruning.

Without this natural pruning mechanism, CommitmentVals history
accumulates at step boundaries (~5.5h on Polygon), growing from 0.7GB
to 6.8GB and causing commit times to degrade from ~3s to 12s+.

Use 60s prune timeout (>=1min triggers PruneSmallBatches adaptive batch
ramp-up, bypassing the SpaceDirty early-exit guard) to drain commitment
history when step boundaries are crossed. This produces a sawtooth
pattern: commits degrade gradually then reset after pruning, with an
overall median of ~5s over 8h of production observation.

Author: madumas

execution/stagedsync/stage_execute.go

@@ -489,6 +489,12 @@ func PruneExecutionStage(s *PruneState, tx kv.RwTx, cfg ExecuteBlockCfg, ctx con
 				"externalTx", useExternalTx,
 			)
 		}
+	} else if cfg.syncCfg.UseForkchoiceFinality {
+		// Forkchoice finality keeps the node stable at tip, eliminating the natural
+		// syncToTip fallbacks that triggered initialCycle=true with aggressive pruning.
+		// Use aggressive timeout (>=1min triggers adaptive batch ramp-up in PruneSmallBatches)
+		// to drain accumulated commitment history at step boundaries.
+		pruneTimeout = 60 * time.Second
 	}
 
 	pruneSmallBatchesStartTime := time.Now()