Cache per-VM SSH sessions

[ricab]

Cache one SSHSession in each VM to run commands in. Mounts are not included

I observe a 3x-5x speedup in info and ls with this.

Suggestions to test disconnection:

• multipass ls after:
  • VM suspend/resume directly on backend (e.g. LXD), with and without a multipass ls in between
  • VM restart directly on backend
  • kill corresponding sshd directly in VM (find it with sudo netstat -tpna — it's the one that doesn't match echo $SSH_CONNECTION)

[codecov[bot]]

Codecov Report

Attention: Patch coverage is 88.42975% with 28 lines in your changes are missing coverage. Please review.

Project coverage is 88.89%. Comparing base (0270a47) to head (0b98ab5). Report is 8 commits behind head on main.

Files	Patch %	Lines
src/daemon/daemon.cpp	38.88%	11 Missing :warning:
...rc/platform/backends/qemu/qemu_virtual_machine.cpp	75.75%	8 Missing :warning:
...tform/backends/libvirt/libvirt_virtual_machine.cpp	73.68%	5 Missing :warning:
.../platform/backends/shared/base_virtual_machine.cpp	96.55%	3 Missing :warning:
...rc/platform/backends/shared/base_virtual_machine.h	0.00%	1 Missing :warning:

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #3431 +/- ## ========================================== + Coverage 88.68% 88.89% +0.20% ========================================== Files 254 254 Lines 14037 14081 +44 ========================================== + Hits 12449 12517 +68 + Misses 1588 1564 -24 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[ricab]

I looks like coverage criteria are different for draft (it was passing before). For the reviewer, notice that:

• The lines that are not covered are essentially lines that I only changed to add/remove a parameter, call a moved function, etc. They were not covered before either (let me know if I am overlooking something).
• I did cover some of those cases anyway, which results in this actually bumping coverage a little bit. See "Additional details and impacted files" in the Codecov post above.

That said, I can try to cover some more if we judge it necessary.

[ricab]

Hey @townsend2010, I have erased those const changes from history :wink:

[townsend2010]

Ok, merging by hand now since the merge queue failed as expected.