Index Cursor, opencode and Gemini CLI sessions

[devcodes9]

Stacked on #55 (the agsearch skill). Base is feat/agsearch-skill, so this diff is the
adapters alone. It retargets to main automatically when #55 merges.

Closes #40.

agsearch indexed Claude Code and Codex. It now indexes Cursor, opencode and Gemini CLI too,
in the same list, with each session resumed by the tool that created it.

Agent	Read from	Resumed with
Claude Code	~/.claude/projects/**/*.jsonl	claude --resume <id>
Codex	~/.codex/sessions/**/*.jsonl	codex resume <id>
Cursor	~/.cursor/chats/**/store.db	cursor-agent --resume <id>
opencode	~/.local/share/opencode/opencode.db	opencode run --session <id>
Gemini CLI	~/.gemini/tmp/**/chats/*.json	gemini --session-file <path>

Why it needed more than a new list entry

Adding a harness was meant to be one line. Five places decided independently what a source
was, so an unknown one was discovered as .jsonl, parsed as Claude, labelled cc, and
resumed with claude --resume. Two of those tables had already drifted (codex against
cx). They now read one record per harness, so a new agent is a parser plus one entry.

Two assumptions also had to go:

• One transcript file, one session. opencode keeps every session in a single database.
  The indexer took the first row's id as the id for the whole file, making every other
  session unreachable. It now registers each session a file contains, and reading one
  filters to it. No change for the four harnesses that write a file per session.
• Every transcript is JSONL. Gemini writes one JSON object per session, so it was
  invisible regardless of what the source table said. The filename test now comes from the
  source record.

Notes per harness

• Cursor stores chats as SQLite. Message blobs are JSON beside binary merkle nodes and
  images, so the scan filters on the leading byte in SQL and opens the store read-only.
  Blob order is insertion order and per-message times were never recorded, so each row
  carries the session's updatedAtMs rather than inventing timestamps. Injected
  <user_info> context is stripped from user turns.
• opencode keeps text in part rows with the role on the parent message. Only text
  parts are indexed; reasoning joins them under --thinking.
• Gemini --resume takes a project-scoped index number, which is not a stable handle
  for a session found by search, so resume goes through --session-file. Gemini records a
  projectHash and no path, so cwd comes from the sibling .project_root.

Verification

Cursor and opencode were tested against real sessions on disk: found by content from the
middle of a conversation, previewed in isolation, and resumed with the right command.

Gemini is covered by fixtures only. The CLI returns IneligibleTierError on a free-tier
account, so no session could be generated to test against, and the parser is written to the
confirmed on-disk schema.

Held-out ranking over 268 queries is unchanged by the larger corpus: 0.504 to 0.507 (@1).
169 tests pass. Cache format bumps to 7 and reindexes once on first run.