Keep VM snapshots local to a test

[comps]

Currently, a test can leave a VM snapshot of a running VM behind (disk+RAM snapshot).

This has the advantage of it being quickly reused by subsequent test re-runs or runs of other tests, but there are several issues:

• If somebody starts up the VM using ./contest-sshvm, it desyncs (corrupts) both the disk and RAM snapshot, which a subsequent test rerun might not catch
• It requires re-reading (re-discovering) IP address of the VM in a new test with a fresh class Guest instance
• It might break SSH if a test tries to run (reuse the snapshot) after NETWORK_EXPIRY hours (7 days)

But mainly, keeping the snapshot around is probably not good design - it keeps the VM in a "fragile" unstable state, which is also a problem for CONTEST_LEAVE_GUEST_RUNNING.

A better solution would be to always keep the VM shut down and without snapshots, but installed-and-ready, between tests.

A test could the easily

1. boot the VM up (1min?)
2. wait for boot to settle (30sec?)
3. Create a persistent ssh connection using ControlMaster
4. create a disk+RAM snapshot
5. do testing by (perhaps repeatedly) restoring the snapshot, using the persistent ssh connection
6. destroy / remove the disk and RAM snapshots

This would also allow hand-tuning of the VM environment for debugging - the test would always re-use an installed VM (given the right tag, eg. Guest() argument), so a human can add some debug/monitoring before rerunning the test. Kind of like manually set-up CONTEST_CONTENT with hand-merged PRs, etc.

[comps]

This would also allow hand-tuning of the VM environment for debugging

It would also allow incremental updates of the VM for ad-hoc fast /per-rule testing, ie. by syncing changed parts of modified content (built tests) to the VM before (re)starting the test.