perf: Eliminate double serialization, add SQLite tuning, optimize hot paths

11 isomorphic performance fixes from deep audit (no behavior changes):

- Eliminate double serialization: store_payload now accepts pre-serialized
  bytes (&[u8]) instead of re-serializing from serde_json::Value. Uses
  Cow<[u8]> for zero-copy when compression is disabled.
- Add SQLite cache_size (64MB) and mmap_size (256MB) pragmas
- Replace SELECT-then-INSERT label upserts with INSERT...ON CONFLICT
  RETURNING in both issues.rs and merge_requests.rs
- Replace INSERT + SELECT milestone upsert with RETURNING
- Use prepare_cached for 5 hot-path queries in extractor.rs
- Optimize compute_list_hash: index-sort + incremental SHA-256 instead
  of clone+sort+join+hash
- Pre-allocate embedding float-to-bytes buffer with Vec::with_capacity
- Replace RandomState::new() in rand_jitter with atomic counter XOR nanos
- Remove redundant per-note payload storage (discussion payload contains
  all notes already)
- Change transform_issue to accept &GitLabIssue (avoids full struct clone)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

src/ingestion/issues.rs

@@ -196,8 +196,8 @@ fn process_single_issue(
     let now = now_ms();
 
     // Transform issue first (outside transaction - no DB access)
-    let payload_json = serde_json::to_value(issue)?;
-    let transformed = transform_issue(issue.clone())?;
+    let payload_bytes = serde_json::to_vec(issue)?;
+    let transformed = transform_issue(issue)?;
     let issue_row = &transformed.issue;
 
     // Wrap all DB operations in a transaction for atomicity
@@ -207,7 +207,7 @@ fn process_single_issue(
         config,
         project_id,
         issue,
-        &payload_json,
+        &payload_bytes,
         issue_row,
         &transformed.label_names,
         &transformed.assignee_usernames,
@@ -226,7 +226,7 @@ fn process_issue_in_transaction(
     config: &Config,
     project_id: i64,
     issue: &GitLabIssue,
-    payload_json: &serde_json::Value,
+    payload_bytes: &[u8],
     issue_row: &crate::gitlab::transformers::IssueRow,
     label_names: &[String],
     assignee_usernames: &[String],
@@ -242,7 +242,7 @@ fn process_issue_in_transaction(
             project_id: Some(project_id),
             resource_type: "issue",
             gitlab_id: &issue.id.to_string(),
-            payload: payload_json,
+            json_bytes: payload_bytes,
             compress: config.storage.compress_raw_payloads,
         },
     )?;
@@ -332,33 +332,27 @@ fn process_issue_in_transaction(
 }
 
 /// Upsert a label within a transaction, returning its ID.
+/// Uses INSERT...ON CONFLICT...RETURNING for a single round-trip.
 fn upsert_label_tx(
     tx: &Transaction<'_>,
     project_id: i64,
     name: &str,
     created_count: &mut usize,
 ) -> Result<i64> {
-    // Try to get existing
-    let existing: Option<i64> = tx
-        .query_row(
-            "SELECT id FROM labels WHERE project_id = ? AND name = ?",
-            (project_id, name),
-            |row| row.get(0),
-        )
-        .ok();
+    let id: i64 = tx.query_row(
+        "INSERT INTO labels (project_id, name) VALUES (?1, ?2)
+         ON CONFLICT(project_id, name) DO UPDATE SET name = excluded.name
+         RETURNING id",
+        (project_id, name),
+        |row| row.get(0),
+    )?;
 
-    if let Some(id) = existing {
-        return Ok(id);
+    // If the rowid matches last_insert_rowid, this was a new insert
+    if tx.last_insert_rowid() == id {
+        *created_count += 1;
     }
 
-    // Insert new
-    tx.execute(
-        "INSERT INTO labels (project_id, name) VALUES (?, ?)",
-        (project_id, name),
-    )?;
-    *created_count += 1;
-
-    Ok(tx.last_insert_rowid())
+    Ok(id)
 }
 
 /// Link an issue to a label within a transaction.
@@ -371,12 +365,13 @@ fn link_issue_label_tx(tx: &Transaction<'_>, issue_id: i64, label_id: i64) -> Re
 }
 
 /// Upsert a milestone within a transaction, returning its local ID.
+/// Uses RETURNING to avoid a separate SELECT round-trip.
 fn upsert_milestone_tx(
     tx: &Transaction<'_>,
     project_id: i64,
     milestone: &MilestoneRow,
 ) -> Result<i64> {
-    tx.execute(
+    let local_id: i64 = tx.query_row(
         "INSERT INTO milestones (gitlab_id, project_id, iid, title, description, state, due_date, web_url)
          VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?7, ?8)
          ON CONFLICT(project_id, gitlab_id) DO UPDATE SET
@@ -385,7 +380,8 @@ fn upsert_milestone_tx(
              description = excluded.description,
              state = excluded.state,
              due_date = excluded.due_date,
-             web_url = excluded.web_url",
+             web_url = excluded.web_url
+         RETURNING id",
         (
             milestone.gitlab_id,
             project_id,
@@ -396,12 +392,6 @@ fn upsert_milestone_tx(
             &milestone.due_date,
             &milestone.web_url,
         ),
-    )?;
-
-    // Get the local ID (whether inserted or updated)
-    let local_id: i64 = tx.query_row(
-        "SELECT id FROM milestones WHERE project_id = ? AND gitlab_id = ?",
-        (project_id, milestone.gitlab_id),
         |row| row.get(0),
     )?;