fix(llm,compression): bound cross-turn Responses chaining and make compression stick

In store mode every user turn chained onto the previous response, so the
provider's stored transcript grew without bound (measured: 889k prompt
tokens for a 37k-token saved history) while every local guard, the
compression pipeline included, measured the saved history. Each chained
tool round also re-sent the system message, which the server appends rather
than dedupes, and a saved compression point was applied exactly once, in the
turn that made it.

Chaining is now bounded. A turn starts from the saved history when the
previous turn's reported prompt reached the chain budget (default: the
model's context window), when the conversation was compressed after that
turn was produced, or when OPENAI_RESPONSES_CHAIN_ACROSS_TURNS is off.
Chained rounds omit an unchanged system head (hash carried in the persisted
Responses state). truncation="auto" is available behind a setting as a
backstop against a chain that outgrows the model's window.

Compression: a saved point is applied at every turn start; the threshold
counts the summary plus the queries after the point instead of the raw
history; re-compression summarises only the tail on top of the last point;
the mid-execution path marks itself persisted and resets the provider chain
so the rebuilt messages are the context; an empty summary is rejected; the
visible "[Context Compression Summary]" rows are no longer replayed as
history; appending the same point twice is a no-op.

Cache hints: a per-user prompt_cache_key and an optional
prompt_cache_retention on Responses API calls.

Measured on Azure with the same client shape as production (stateless
OpenAI client, server-side tools, PDF part): tokens billed on the sixth turn
fell from 58k to 35k, tool rounds add tens of tokens instead of ~2.8k, the
turn after a compression reused the saved summary in under two seconds
instead of re-summarising, and the round after a mid-execution compression
started from the compressed context instead of the full stored transcript.

docs/content/Models/cloud-providers.mdx

@@ -66,6 +66,9 @@ For OpenAI models that support it, DocsGPT can call the newer **Responses API**
 
 - `reasoning_effort` — per-model reasoning effort hint (for example `medium`) declared in the model catalog.
 - `OPENAI_RESPONSES_STORE` (default `false`) — when `true`, lets OpenAI persist Responses API state server-side and chains calls with `previous_response_id`. When `false`, DocsGPT requests encrypted reasoning items and persists those ciphertext items itself so reasoning survives tool calls and later conversation turns without OpenAI retaining the response.
+- `OPENAI_RESPONSES_CHAIN_ACROSS_TURNS` (default `true`) and `OPENAI_RESPONSES_CHAIN_BUDGET_TOKENS` (default: the model's `context_window`) — in store mode a new user turn chains onto the previous response, which keeps the provider's stored transcript and its prompt cache warm. The chain is bounded: once the previous turn's reported prompt reaches the budget, or after the conversation history has been compressed, the next turn starts from DocsGPT's own saved history instead. Chained tool rounds do not re-send an unchanged system message.
+- `OPENAI_RESPONSES_TRUNCATION_AUTO` (default `false`) — send `truncation: "auto"` so the provider trims middle items instead of failing a request that exceeds the model's window.
+- `OPENAI_PROMPT_CACHE_KEY` (default `true`) and `OPENAI_PROMPT_CACHE_RETENTION` (default unset) — prompt-cache hints on Responses API calls: a per-user `prompt_cache_key`, and a `prompt_cache_retention` value such as `24h` where the provider supports it.
 - `V1_SESSION_TTL_SECONDS` (default `86400`) — lifetime of the hashed Redis mapping used to associate OpenAI-compatible client session headers with hidden DocsGPT conversations. Explicit DocsGPT conversation IDs always take precedence.
 
 See [App Configuration](/Deploying/DocsGPT-Settings) for the full settings reference.