ci: move fuzz corpus to S3

[jouho]

Resolved issues:

Part of #1447

Description of changes:

• This change enables the upload of corpus files generated from fuzz tests to an S3 bucket and allows downloading them to start a new fuzz test using the accumulated corpus files.

• A new environment variable, CORPUS_UPLOAD_LOC, has been introduced. When this variable is specified, the fuzz test will download corpus files from the specified S3 bucket for a new fuzz test. After the test, any newly generated corpus files are uploaded to the corresponding S3 location, allowing continuous improvement of the corpus.

• Not a change in this PR, but new policy s2n-tls-fuzz-corpus-s3-access is attached to corresponding CodeBuild project to grant access to S3 bucket storing corpus files.

Call-outs:

This feature can be turned on or off by specifying CORPUS_UPLOAD_LOC in the CodeBuild configuration by overriding the environment variable. I suggest first turning it on for s2nFuzzScheduled (a daily job that runs a single fuzz test for ~8 hours) to ensure long-term viability. Afterward, we can enable it for s2nFuzzBatch, which runs against each PR.

There are also some other tasks left to do after this change:

• Currently, the daily task is only running 6 fuzz tests out of ~30 different tests. We should look to activate the inactive ones.
• Artifacts are currently stored and overwritten in one S3 location. We should store them in separate locations with timestamp on it for potential debugging.
• After enabling this feature for s2nFuzzBatch, we should probably remove all the cluttering corpus files stored in the s2n-tls repo.

Testing:

I have some test runs for both s2nFuzzBatch and s2nFuzzScheduled using this PR as source. You can see their build log:

• s2nFuzzBatch (there are some failed builds when I was configuring correct access policy to s3 bucket).
• s2nFuzzScheduled

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

[dougch]

thanks for getting this done