fix: move pre_llm_call plugin context to user message, preserve prompt cache (#5146)

Plugin context from pre_llm_call hooks was injected into the system
prompt, breaking the prompt cache prefix every turn when content
changed (typical for memory plugins). Now all plugin context goes
into the current turn's user message — the system prompt stays
identical across turns, preserving cached tokens.

The system prompt is reserved for Hermes internals. Plugins
contribute context alongside the user's input.

Also adds comprehensive documentation for all 6 plugin hooks:
pre_tool_call, post_tool_call, pre_llm_call, post_llm_call,
on_session_start, on_session_end — each with full callback
signatures, parameter tables, firing conditions, and examples.

Supersedes #5138 which identified the same cache-busting bug
and proposed an uncached system suffix approach. This fix goes
further by removing system prompt injection entirely.

Co-identified-by: OutThisLife (PR #5138)

hermes_cli/plugins.py

@@ -441,8 +441,18 @@ class PluginManager:
         plugin cannot break the core agent loop.
 
         Returns a list of non-``None`` return values from callbacks.
-        This allows hooks like ``pre_llm_call`` to contribute context
-        that the agent core can collect and inject.
+
+        For ``pre_llm_call``, callbacks may return a dict describing
+        context to inject into the current turn's user message::
+
+            {"context": "recalled text..."}
+            "recalled text..."          # plain string, equivalent
+
+        Context is ALWAYS injected into the user message, never the
+        system prompt.  This preserves the prompt cache prefix — the
+        system prompt stays identical across turns so cached tokens
+        are reused.  All injected context is ephemeral — never
+        persisted to session DB.
         """
         callbacks = self._hooks.get(hook_name, [])
         results: List[Any] = []