Fix gateway kernel shutdown

[kevin-bates]

While working on the next Enterprise Gateway release, I ran into a couple of issues encountered when shutting down Jupyter Server when a Gateway is configured and there are outstanding kernel instances.

1. The loop walking the set of outstanding kernels was being modified producing "RuntimeError: dictionary changed size during iteration".
2. Attempts to delete a kernel known to Server but unknown to Gateway (due to culling or inadvertent termination on the gateway) was triggering an exception for the 404 when, for shutdown operations, this can be ignored.

This pull request addresses these issues and adds an appropriate test as well. (I confirmed expected test failures prior to applying the fixes.)

[codecov-commenter]

Codecov Report

Merging #874 (528ab42) into main (29911b1) will increase coverage by 0.07%.
The diff coverage is 61.53%.

@@            Coverage Diff             @@
##             main     #874      +/-   ##
==========================================
+ Coverage   70.18%   70.25%   +0.07%     
==========================================
  Files          65       65              
  Lines        7680     7683       +3     
  Branches     1283     1284       +1     
==========================================
+ Hits         5390     5398       +8     
+ Misses       1902     1897       -5     
  Partials      388      388              

Impacted Files	Coverage Δ
jupyter_server/gateway/managers.py	59.79% <61.53%> (+1.87%)	⬆️
jupyter_server/services/kernels/kernelmanager.py	78.31% <0.00%> (-0.33%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 29911b1...528ab42. Read the comment docs.

[blink1073]

Thanks!

[blink1073]

@meeseeksdev please backport to 1.x