Improvements for CLI load time

[chrisjsewell]

Thanks for the clarification, it was just since you asked for feedback and I didn't understand that the time difference has disappeared. If the time difference is negligible I'm ok with the merge.
Just to report, I also run on my computer and I get:

This branch:

Benchmark #1: verdi
  Time (mean ± σ):     387.7 ms ±  18.9 ms    [User: 315.7 ms, System: 65.9 ms]
  Range (min … max):   371.0 ms … 428.8 ms    10 runs

develop:

Benchmark #1: verdi
  Time (mean ± σ):     338.6 ms ±   5.1 ms    [User: 275.4 ms, System: 57.6 ms]
  Range (min … max):   331.4 ms … 346.4 ms    10 runs

So I'm OK to merge.

However, while looking into it with cProfile+snakeviz, I realised that a call to cmd_computer was taking a lot of time (~80ms) even when the command was not involved (e.g. when expanding "verdi data bands sho").

In this branch, if you comment out the whole content of the cmd_computer.py file, indeed the time goes down significantly from ~370ms to ~305ms.
If you just uncomment this single line:

from aiida.transports import cli as transport_cli

Then the time goes up again.

If you want to give a look into this already now, great, otherwise I'm ok with merging this as is since this would be a further improvement, but in this case could you please open an issue mentioning this? (I also have the feeling that there might be at least one or two similar things one could improve, e.g. also cmd_archive seems to be taking some time, but the cProfile+snakeviz is quite complex to "parse" and I still don't know if it's in addition to the time of cmd_computer or not).

Originally posted by @giovannipizzi in #5058 (comment)