feat(nvd): retry 429ed records after conversion is completed

[jess-lowe]

Call me hopeful, but we should attempt to convert the 429ed records again.

this maintains a threadsafe list of records that 429ed and then retries them at the end of conversion.

• Updated RepoTags so that temporary 429 (rate limit / too many requests) errors do not mark a repository as permanently invalid in the repository tags cache (SetInvalid).
• Updated the worker pool and worker function to collect any CVE records that fail with models.Error (rate limits) into a thread-safe retryList. After the initial processing pass completes, a secondary worker pool is launched to attempt converting those rate-limited records again.

[another-rex]

Let's just do the first point. And rather than not setting it as invalid in the cache, set it as invalid, just with a shorter timeout than the standard 1 week. (Maybe just 1 hour is good enough, just so we are not spamming requests when multiple records relies on 1 429 host)

It'll run again in 12 hours and pick it back up.