Add improvements to recent tensorflow-tensorboard fix

[joshuacwnewton]

Checklist

GitHub

• I've given this PR a concise, self-descriptive, and meaningful title
• I've linked relevant issues in the PR body
• I've applied the relevant labels to this PR
• I've assigned a reviewer

PR contents

• I've consulted SCT's internal developer documentation to ensure my contribution is in line with any relevant design decisions
• I've added relevant tests for my contribution N/A
• I've updated the relevant documentation for my changes, including argparse descriptions, docstrings, and ReadTheDocs tutorial pages N/A

Description

PR #3034 was included in #3041 to fix #3035. But, on closer inspection, I noticed some issues:

• The fix was only applied to the dev install, not the release install.
  • Solution: Use a variable to store the name of the requirements file, then have the pip install be done in a single command for both types of installation.
• The fix used multiple separate commands, which invalidates the $? status check.
  • Solution: Use && to group commands, so that $? remains a valid check.
  • "Commands separated by a double ampersand && are to be run synchronously, with each one running only if the last did not fail (a fail is interpreted as returning a non-zero return status)."
• The fix missed a comment that had been mentioned in Bump ivadomed version to 2.5.0 #3034 but left out of Support of multichannel and multiclass models #3041.

Linked issues

Fixes #3035.

[lrouhier]

It looks good to me! Thank you for catching that, It makes it more elegant!