Update GitHub Actions action versions

[jwodder]

We are currently using older versions of the updated GitHub Actions actions, and these older versions still use Node 12, which is deprecated.

This PR also configures Dependabot to create PRs updating workflows whenever a new version of a used action is released, as discussed in datalad/datalad-extensions#105.

[codecov]

Codecov Report

Base: 74.88% // Head: 76.01% // Increases project coverage by +1.12% 🎉

Coverage data is based on head (dccc798) compared to base (18f0bf9).
Patch has no changes to coverable lines.

Additional details and impacted files

@@            Coverage Diff             @@
##            maint    #7082      +/-   ##
==========================================
+ Coverage   74.88%   76.01%   +1.12%     
==========================================
  Files         355      355              
  Lines       59101    59101              
  Branches     6318     6318              
==========================================
+ Hits        44260    44925     +665     
+ Misses      14827    14162     -665     
  Partials       14       14              

Impacted Files	Coverage Δ
datalad/_version.py	45.68% <0.00%> (-0.28%)	⬇️
datalad/__init__.py	86.00% <0.00%> (+2.50%)	⬆️
datalad/tests/utils.py	51.98% <0.00%> (+20.65%)	⬆️
datalad/tests/test_tests_utils.py	68.55% <0.00%> (+68.55%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[yarikoptic]

hm, I dewarped the changelog snippet and then looked back. benchmarks are failing with something which might actually relate:

fatal: HEAD^1: not a valid SHA1
ram [7110660]: 
Error: Process completed with exit code 128.

rerunning : https://github.com/datalad/datalad/actions/runs/3236005986

[yarikoptic]

seems to be consistently failing, @jwodder please check what could be the difference between two versions of checkout which would affect it

[jwodder]

@yarikoptic The Benchmark workflow was previously using v1 of actions/checkout, which checks out the complete history by default, yet v2 and up only check out a single commit by default. Setting fetch-depth: 0 should fix it.

[yarikoptic]

@yarikoptic The Benchmark workflow was previously using v1 of actions/checkout, which checks out the complete history by default, yet v2 and up only check out a single commit by default. Setting fetch-depth: 0 should fix it.

oh... then it is something to keep for every installation of any module which uses versioneer, versioningit and alike or version would not be deducable. That lead to tests failure on OSX, and it seems even more "fun" situation on testing extensions which then lead to "reinstallation" of datalad from pypi (instead of already reusing the one from checkout) while installing extension: from https://github.com/datalad/datalad/actions/runs/3237400351/jobs/5304414220

2022-10-12T19:51:20.2161287Z Collecting datalad>=0.17.6
2022-10-12T19:51:20.2447327Z   Downloading datalad-0.17.6-py3-none-any.whl (1.4 MB)

so I added to those and other workflows similar depth setting now and pushed. I added even to workflows where it is not exactly needed (e.g. crippled fs - we do not run tests on version to be ok, or disabled windows) so we do not need to figure it out later.
I guess it is something to keep attention to (well, I have merged most of your PRs by now ;))

[yarikoptic]

I have checked logs of one of the extensions and datalad reported to be

version: 0.17.6+62.g0d3f8e8ed

so all good now, let's proceed, but now with a better awareness of change in semantic of new version of checkout

[yarikoptic-gitmate]

PR released in 0.17.7