Add Cache-Control to the commit API route (#8580)

Stack from [ghstack](https://github.com/ezyang/ghstack/tree/0.14.0)
(oldest at bottom):
* #8584
* #8582
* __->__ #8580
* #8579
* #8578
* #8576

**Impact:** HUD commit page consumers (CDN-cached reads)
**Risk:** low

## What
Set a short-TTL `Cache-Control` header (`s-maxage=60,
stale-while-revalidate=240`) on the
`/api/[repoOwner]/[repoName]/commit/[sha]` endpoint, which previously
sent no caching headers.

## Why
The commit endpoint calls into GitHub via `fetchCommit`, and every
repeat view hit the origin, adding to the shared PyTorchBot
installation's rate-limit budget. Caching at the CDN collapses repeat
views of the same sha to at most one origin call per minute. The TTL is
kept short because the response mixes immutable commit metadata with
live CI job data that must stay fresh.

Signed-off-by: Jean Schmidt <contato@jschmidt.me>

Author: jeanschmidt

torchci/pages/api/[repoOwner]/[repoName]/commit/[sha].ts

@@ -19,15 +19,14 @@ export default async function handler(
 ) {
   const workflowId = parseInt(req.query.workflowId as string, 10) || 0;
   const runAttempt = parseInt(req.query.runAttempt as string, 10) || 0;
-  res
-    .status(200)
-    .json(
-      await fetchCommit(
-        req.query.repoOwner as string,
-        req.query.repoName as string,
-        req.query.sha as string,
-        workflowId,
-        runAttempt
-      )
-    );
+  const data = await fetchCommit(
+    req.query.repoOwner as string,
+    req.query.repoName as string,
+    req.query.sha as string,
+    workflowId,
+    runAttempt
+  );
+  // Short TTL because the response bundles live CI job data with the immutable commit metadata.
+  res.setHeader("Cache-Control", "s-maxage=60, stale-while-revalidate=240");
+  res.status(200).json(data);
 }