Fix multi process kubeconfig access support

[michaelbeaumont]

Description

Multi process support was broken on every platform, just even worse on windows.

The problem on Windows was that by locking the kubeconfig file itself, client-go/tools/clientmd.ModifyConfig is unable to write to the file.
The intended use of flock is to create a lock file to protect some critical section.
This is also what client-go does internally (hence the .eksctl prefix for the lock file).

Additionally, we must use the the same flock.Flock to lock and unlock the file. Doing otherwise leads to deadlocks and concurrent writes.

This wasn't caught because the test that was written with the PR didn't actually test anything, it just made a bunch of calls to os.Write, not kubeconfig.Write.

Fixes #2376

Checklist

• Added tests that cover your change (if possible)
• Added/modified documentation as required (such as the README.md, or the userdocs directory)
• Manually tested
• Added labels for change area (e.g. area/nodegroup), target version (e.g. version/0.12.0) and kind (e.g. kind/improvement)
• Make sure the title of the PR is a good description that can go into the release notes

[cPu1]

Looks like the library was also being used incorrectly. Instead of unlocking the existing file handle, it was creating a new Flock instance and trying to unlock it (which is a no-op).

[michaelbeaumont]

Looks like the library was also being used incorrectly. Instead of unlocking the existing file handle, it was creating a new Flock instance and trying to unlock it (which is a no-op).

Yes exactly, that was broken on every platform