fix(kernel,server,workflow): per-stage retry budget, sane freestyle budget, tail /events

Three defects surfaced by a live headless freestyle_planning run (all made a
long execution workflow fail or stall in ways invisible to a REST operator):

1. retryCount was session-global (DefaultOrchestrationReducer): TransitionExecuted
   updated currentStageId but never reset retryCount, while shouldRetry gates on
   it per-stage. An early stage that burned its 3 retries starved every later
   stage of its own — the next stage's first *retryable* failure went terminal.
   Reset retryCount on stage entry; back-edge loops stay bounded by the separate
   refinementIterations counter. Regression test added.

2. Freestyle stages ran at StageConfig's 4096-token default (ExecutionPlanCompiler
   never set a budget) vs 16384-32768 for static role_pipeline stages. A tool-heavy
   stage truncated its context every round, evicting the file_read results it just
   gathered, so it re-read forever and never accumulated enough to write. Set 16384.

3. GET /events returned the OLDEST `limit` events (readFrom(0) then take): for a
   session past `limit` events the tail — including a pending ApprovalRequested —
   was invisible to a headless/REST poller, the exact caller POST /approve serves.
   Default now tails; explicit fromSeq still paginates forward.

core/kernel/src/main/kotlin/com/correx/core/kernel/orchestration/DefaultOrchestrationReducer.kt

@@ -53,7 +53,13 @@ class DefaultOrchestrationReducer : OrchestrationReducer {
             pauseReason = null,
         )
 
-        is TransitionExecutedEvent -> state.copy(currentStageId = p.to)
+        // Entering a new stage resets the retry budget: retryCount is per-stage (RetryAttemptedEvent
+        // is stage-scoped, maxAttempts is a per-stage cap), so without this reset the budget is shared
+        // session-wide — an early stage that exhausts its retries starves every later stage of its own,
+        // turning the next stage's first retryable failure into a terminal WorkflowFailed (live repro:
+        // implement burned 3/3, then review's retryable JSON-decode error got 0 retries). Back-edge
+        // loops stay bounded by the separate refinementIterations counter, so this can't loop forever.
+        is TransitionExecutedEvent -> state.copy(currentStageId = p.to, retryCount = 0)
 
         is RetryAttemptedEvent -> state.copy(
             status = OrchestrationStatus.RUNNING,