#1682: fix help tests

[Ifropc]

Part of #1682

In this PR help tests are fixed.
As part of the change, I re-organized build_host_function_parameters to return either previous result or help message. All callers up the stack now respect this behavior.

Why

Previously, we were using clap's get_matches_from that exits with code 0 (printing help message). This made our tests exit with code 0 as well (when calling --help).
As part of the change we handle DisplayError returning help message, and keeping existing behavior for all other clap errors.

Known limitations

Some of the tests are now failing in CI. They were previously effectively ignored, as the whole suite was exiting as soon as help tests were running.
This tests are fixed in other PRs (see #1682 )

[leighmcculloch]

This tests are fixed in other PRs (see #1682 )

What needs to merge before this PR can be merged?

[Ifropc]

@leighmcculloch we can either merge all listed PRs at the same time or merge all of them here first

• #1682: fix help tests #1734
• #1682: fix build tests #1736
• 1682 fix config/version tests #1737
• 1682 fix init test #1738

[elizabethengelman]

This is looking great! Thanks for digging into it - it was driving me crazy ;p

we can either merge all listed PRs at the same time or merge all of them here first

W/r/t when to merge this and the other PRs, i like the idea of merging those test fixing into this PR first.

[leighmcculloch]

Thanks for figuring out why these tests were preventing from running. Great find and learning for us all.

For the solution, I'd like us to reconsider the approach here, see the inline comment. Please lmk if there's a reason we can't do the error approach.

[leighmcculloch]

One ask, otherwise this looks good, very easy to follow.

[leighmcculloch]

I think we need to do this at the top level, rather than inside the arg parsing, otherwise the exit codes will be incorrect when generating the help when no --help is passed. It's a small thing, but also I believe a small fix to keep passing the error as is up the chain, and at the top level doing as you do here.

[leighmcculloch]

Same comment about doing this at the top level.