Conditionally run set-variables job in run-profiling.yaml workflow file

[matt-graham]

At the moment the set-variables job in .github/workflows/run-profiling.yaml runs unconditionally:

TLOmodel/.github/workflows/run-profiling.yaml

Lines 31 to 46 in b9e157f

	set-variables:
	name: Create unique output file identifier and artifact name
	runs-on: ubuntu-latest
	outputs:
	profiling-filename: ${{ steps.set-profiling-filename.outputs.name }}
	artifact-name: $${{ steps.set-artifact-name.outputs.name }}
	steps:
	- id: set-profiling-filename
	name: Set profiling output file name
	run: |
	echo "name=${GITHUB_EVENT_NAME}_${GITHUB_RUN_NUMBER}_${GITHUB_SHA}" >> "${GITHUB_OUTPUT}"
	- id: set-artifact-name
	name: Set artifact name
	run: |
	echo "name=profiling_results_${GITHUB_RUN_NUMBER}" >> "${GITHUB_OUTPUT}"

which means even if the subsequent steps are skipped we get a workflow success status rather than skipped status, which for example stops us using a query like https://github.com/UCL/TLOmodel/actions/workflows/run-profiling.yaml?query=is%3Asuccess to find successful runs of the workflow (or at least ones in which any profiling was performed!).

I think if we add an if condition something like

if: (github.event_name != 'issue_comment') || ((github.event_name == 'issue_comment') && (github.event.comment.body == '/run profiling')

we should prevent the job being run when the succeeding jobs will be skipped, though then we do end up needing to keep the condition here in sync with the logic in the next jobs.

[willGraham01]

I feel like the if condition is the lesser of two evils - the alternative (which I think we tried when first setting up the profiling) is to put all of the logic in the run-profiling workflow into the command argument to the comment-triggered stuff, which wasn't the most readable.