feat(executor): upgrade to Snakemake v7.32.4

[giuseppe-steduto]

Amend the overridden executor to reflect the changes in the new version
of Snakemake, in particular with regard to the change of the
_wait_for_jobs method into a coroutine.

Closes #31
Closes reanahub/reana-client#655

[codecov]

Codecov Report

Attention: 6 lines in your changes are missing coverage. Please review.

Comparison is base (4782678) 3.84% compared to head (4a3f359) 3.82%.

Additional details and impacted files

@@            Coverage Diff            @@
##           master     #81      +/-   ##
=========================================
- Coverage    3.84%   3.82%   -0.03%     
=========================================
  Files           6       6              
  Lines         182     183       +1     
=========================================
  Hits            7       7              
- Misses        175     176       +1     

Files	Coverage Δ
reana_workflow_engine_snakemake/executor.py	0.00% <0.00%> (ø)

[mdonadoni]

Why do we need to move this after the installation of reana-commons?

[giuseppe-steduto]

When building locally in debug mode and installing a version of reana-commons that is only available locally (not released on PyPi yet), this step was failing because it could not find the required reana-commons version.
Moving it after the installation of reana-commons solves the issue.

[mdonadoni]

Any reasons why you needed to change the version of r-commons locally? Just to understand what is the use case

[giuseppe-steduto]

It's just usually what I do when updating r-commons and some other component at the same time, especially because I can quickly check if the changes were propagated correctly by just looking at the installed r-commons version

[tiborsimko]

FWIW it would be nice to keep the same order for all the REANA components, i.e. either make this change everywhere, or just keep the original order here too.

Personally when I was testing your branch, I just downgraded the setup.py line to require 0.9.5 by a local edit, and tested it that way.

IMO the old order was acceptable, because we usually only bump r-commons version when the new r-commons is released on PyPI, so the branch remains in a "easily testable" locally for a long time, so the above setup.py micro-editing is not even needed.

That said, we could also change the order if it make things easier -- and there should be no danger of overriding some dependencies in that way -- but then let's do it for all the components systematically, for the overall consistency, and in a nicely separate commit, for better housekeeping.

[mdonadoni]

After discussing with Tibor, given that the UID of the user can be changed with WORKFLOW_RUNTIME_USER_UID, maybe it's better if we directly create the directory in the Dockerfile with mkdir /.cache/snakemake and then we set the env variable XDG_CACHE_HOME to /.cache? In this way, independently by which user is used, the directory always exists

[tiborsimko]

The above does not work for me. The following does:

diff --git a/Dockerfile b/Dockerfile
index c73f1a2..d8b036a 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -103,7 +103,7 @@ ENV TERM=xterm \
     PYTHONPATH=/workdir

 # Create and set cache directory to be used by Snakemake
-RUN mkdir -p /.cache/snakemake && chmod 644 /.cache
+RUN mkdir -p /.cache/snakemake && chmod ug+rwx /.cache/snakemake
 ENV XDG_CACHE_HOME=/.cache

 # Set image labels