machine: add SSHSession to reuse SSH connections

[inknos]

Each ssh.Dial to localhost costs ~77ms. I counted 6 SSH calls during a machine start, ~464ms is spent on connection init. Some more time can be saved by incorporating bash commands in one.

• Merge the two separate SSH calls (write tmpfiles config + run systemd-tmpfiles) into a single bash -c command joined with &&.

• MountVolumesToVM now opens a single SSHSession for all volumes

• Combine mkdir and mount steps into a single bash -c command per volume, further halving the number of SSH round-trips.

Time measurements are done by running
bin/podman machine start --log-level=debug
and checking the time in the logs (this and the table were done by Claude):

Init (unchanged, not optimized):
+---+--------+--------+---------+------------------------------+
| # | Dial | Run | Total | Purpose |
+---+--------+--------+---------+------------------------------+
| 1 | 95ms | 2156ms | 2251ms | SSH readiness probe |
| 2 | 85ms | 738ms | 823ms | inject proxy env |
+---+--------+--------+---------+------------------------------+

Volume mounts — before (4 connections, 4 dials):
+---+--------+--------+---------+------------------------------+
| # | Dial | Run | Total | Purpose |
+---+--------+--------+---------+------------------------------+
| 1 | 66ms | 341ms | 408ms | vol 1: mkdir |
| 2 | 70ms | 329ms | 400ms | vol 1: mount |
| 3 | 76ms | 423ms | 499ms | vol 2: chattr + mkdir |
| 4 | 72ms | 270ms | 342ms | vol 2: mount |
+---+--------+--------+---------+------------------------------+
| 284ms | 1363ms | 1649ms |
+--------+--------+---------+

Volume mounts — after (1 dial, shared session):
+---+--------+--------+---------+------------------------------+
| # | Dial | Run | Total | Purpose |
+---+--------+--------+---------+------------------------------+
| 1 | 64ms | -- | -- | open shared session |
| 2 | 0ms | 586ms | 586ms | vol 1: mkdir + mount |
| 3 | 0ms | 527ms | 527ms | vol 2: chattr + mkdir + mount|
+---+--------+--------+---------+------------------------------+
| 64ms | 1113ms | 1113ms |
+--------+--------+---------+
Saved: ~220ms dial, ~250ms total (3 fewer connections)

Fixes: https://redhat.atlassian.net/browse/RUN-4470

Checklist

Ensure you have completed the following checklist for your pull request to be reviewed:

• Certify you wrote the patch or otherwise have the right to pass it on as an open-source patch by signing all
  commits. (git commit -s). (If needed, use git commit -s --amend). The author email must match
  the sign-off email address. See CONTRIBUTING.md
  for more information.
• Referenced issues using Fixes: #00000 in commit message (if applicable)
• Tests have been added/updated (or no tests are needed)
• Documentation has been updated (or no documentation changes are needed)
• All commits pass make validatepr (format/lint checks)
• Release note entered in the section below (or None if no user-facing changes)

Does this PR introduce a user-facing change?

None

[packit-as-a-service]

[NON-BLOCKING] Packit jobs failed. @containers/packit-build please check. Everyone else, feel free to ignore.

[inknos]

@Honny1 ptal, I am not sure which tests should I write for this change?

[Honny1]

I don't think we need new tests for these changes. I am adding the No New Tests label.

[Luap99]

Ideally we switch the qemu mounts over to systemd units just like we have on applehv.
For hyperV 9p mounts I think we will always need the manual ssh mounts though.

machine tests are so flaky I agree we cannot really have performance tests so no tests seem fine for this, existing tests should cover these code paths.

[baude]

looks like the qemu failures are legit?

@Luap99 what would you like to do here? Do you want to merge this and go back or just require @inknos to do that change now?

[Luap99]

if we touch the code now we might as well do it right with systemd, I mean the code for that should already be in applehv so just need to share it here in qemu as well

That feels nicer than joining one huge command.

[inknos]

got it, thanks for the feedback!

[baude]

got it, thanks for the feedback!

thanks @inknos !