Fix .aws/config destroying multi-process concurrency bug

[snickell]

Fix multiprocess concurrency bug that trashes ~/.aws files

Prevents the error: Unable to parse config file: where you have to either fix ~/.aws/config or delete ~/.aws/credentials. Suddenly its rock solid!

We previously had a mutex to try to prevent the worst of these, but this only works with threads. If multiple processes try to use aws-google at once, they can simultaneously run aws configure set ... which writes to both ~/.aws/config and ~/.aws/credentials, trashing one or the other.

The solution in this PR is to use an flock on ~/.aws/aws-google.lock, which can synchronize access by multiple processes. To prevent infinite hangs, processes will only block for 60s on the lock before raising an exception.

Because an flock is a kernel level feature on a file descriptor, when the descriptor is destroyed, even if the process ends in a SEGFAULT or whatever, the kernel should delete the lockfile.

Testing

• Was able to verify the issue with moderate consistency (about 2/3 times it corrupted) by launching 20x bin/aws_access runs at once.
• After this PR, I can't get it to corrupt the file.
• Adds tests written by codex, but frankly, I've only skimmed the tests

[snickell]

Hmmm, found what might be an issue, or maybe I'm just using it wrong locally from the source dir? Going draft since I'm on vacation for a bit.