fix: ensure cache is done writing before exit (#6603)
### Description
This PR contains intertwined changes:
- Waiting on all cache writes to finish before we exit the run. We
leverage the signal handler here so we can make sure we wait regardless
of how/why the run exits.
- Printing errors from artifact upload failures. This differs from Go
where we [throw generated errors
away](https://github.com/vercel/turbo/blob/main/cli/internal/cache/async_cache.go#L79C12-L79C12)
(I'm not sure this is desired since we do [format these errors for human
consumption](https://github.com/vercel/turbo/blob/main/cli/internal/client/cache.go#L70))
- Correctly deserialize Vercel API errors. As per the
[docs](https://vercel.com/docs/rest-api/errors) there's a containing
object with an `error` field that has the information.
- Pass team information when saving artifacts. This matches all of the
other artifact api calls.
Each commit is reviewable on it's own.
### Testing Instructions
The `turbo` cli is a great candidate for these. Make sure you're logged
in and linked to Vercel team.
For testing that we wait until upload finishes:
```
# Ensure there's no error printed at the end of the run and the "Finishing writing to cache" message appears while we upload the large debug binary
[0 olszewski@chriss-mbp] /Users/olszewski/code/vercel/turborepo $ turbo_dev build --filter=cli --force
...
...Finishing writing to cache...
# Should be a FULL TURBO due to the previous command
[0 olszewski@chriss-mbp] /Users/olszewski/code/vercel/turborepo $ turbo_dev build --filter=cli --remote-only --remote-cache-timeout=120
```
Closes TURBO-1770
---------
Co-authored-by: Chris Olszewski <Chris Olszewski>
