[Shabonasar]

The [ref: tool<i>turn<n>] label is now the stored name of a tool output: it is saved with the result and restored with the history, and turn numbers continue across runs. Before, the label was shown only while the key was in the current run's registry and every run restarted at turn0, so:

• the restored history lost the labels: the first request of each later turn missed the prompt cache and Claude dropped thinking blocks bound to the old prefix;
• keys repeated between turns: {{tool0turn0}} from an old turn silently resolved to the current turn's output.

Fixes #627

live:     ToolNode key tool0turn7 -> ON_RUN_STEP_COMPLETED.tool_call { output, ref, unresolved_refs? } -> host stores it
restore:  formatAgentMessages: ref -> ToolMessage.additional_kwargs._refKey
          Run: turn counter = highest restored turn + 1   (keys never repeat)
request:  annotateMessagesForLLM labels by _refKey, no registry lookup, after truncation (byte-equal to live)

• An old key is not in the registry, so it stays literal with the existing [unresolved refs: ...] warning.
• Handoff agents share one registry (one seed); subagents have their own run and history.
• Tests that expected "no label when the key is missing from the registry" now expect the label.
• LibreChat needs no change: the fields ride in the stored tool_call part.
• Not changed: a placeholder in the same parallel batch as the call it names stays literal without a warning.

Measurements

Real Anthropic API through LibreChat, 6 turns, adaptive thinking, code interpreter running code, chained bash_tool calls, MCP tool, restart after turn 3, turn 5 references a key from turn 1. "series" = the LibreChat series plus agents #613, #614, #615, #619 (LibreChat-AI/LibreChat#17102); dev = LibreChat dev with agents main. This PR has no dependency on the series and was added on top of each. "before" used a 1h cache TTL, "after" 5m.

model	LibreChat	agents	cache hit	misses	dropped thinking blocks	400
Haiku 5.5	series #17102	before	87.9%	4	28	0
Haiku 5.5	series #17102	this PR	92.9%	0	0 (2nd run: 1, from #618)	0
Sonnet 5.5	series #17102	before	86.1-88.4%	3	10	0
Sonnet 5.5	series #17102	this PR	90.8%	0	0	0
Haiku 5.5	dev	before	88.6%	5	0	0
Haiku 5.5	dev	this PR	87.5%	5	0	0

With the series, keys run tool0turn0...tool0turn8 without repeats, and the turn-5 probe returns [unresolved refs: tool0turn0] instead of a wrong output. On dev the labels are now stable too, but its misses come from other restore differences fixed by the series (LibreChat-AI/LibreChat#17102).

Testing

• Full suite: 6568 passed; the 16 failures also fail without this change (3 llm specs need API keys, 2 handoff/checkpoint specs).
• New ToolNode.persistedReferences.test.ts (9): fields on direct, early and eager completions; none when the feature is off; aggregator -> persistence -> restore gives the live wire content; turn seeding; old key stays unresolved.
• tsc --noEmit, eslint, check:circular-deps clean.

Type of change

• Bug fix

Checklist

• I reviewed my own changes
• Relevant tests have been added or updated
• Existing relevant tests pass
• The change does not introduce new warnings or errors