Recover when a stale daemon pid cannot be killed

[bvinc]

If buckd.info names a pid we cannot kill - it was reused by a process owned by another user, say - the kill error aborted the connect before clean_daemon_dir ran. The stale buckd.info survived, so every later invocation failed the same way and the only way out was deleting the daemon dir by hand.

Report the failed kill and start a new daemon instead of aborting.

This fixes an error seen by a user where after a PID on windows was re-used, the user was stuck until deleting their ~/.buck/buckd directory.

> buck2 killall; buck2 server

  Killed buck2.exe (26140).
  Could not connect to buck2 daemon (Could not connect to daemon), killing daemon..
  Command failed: Failed to connect to buck daemon.
      Daemon process info from C:\Users\...\.buck\buckd\C\git\roblox\game-engine\v2\buckd.info:
      pid: 21076
      endpoint: tcp:62852
      version: baf6de80...
      pid present on system: yes
      Try running `buck2 kill` and your command afterwards.
      Alternatively, try running `rmdir /s /q %USERPROFILE%\.buck\buckd` and your command afterwards
  Caused by:
      0: Error connecting to the daemon, daemon stderr follows:

[meta-codesync]

This pull request has been imported. If you are a Meta employee, you can view this in D114131935. (Because this pull request was imported automatically, there will not be any future comments.)

[scottcao]

Review automatically exported from Phabricator review in Meta.

[meta-codesync]

This pull request has been merged in 8094df3.