[bugfix] Fix pipeline hook override

[jjotero]

This PR fixes a hook attachment leak occurring when a given hook from a base class is overridden by a derived class.

When overriding a pipeline hook, the method in the derived class would still get attached to the pipeline despite that it is not decorated to do so. See the example below:

@rfm.simple_test
class Base(rfm.RunOnlyRegressionTest):
    valid_systems = ['*']
    valid_prog_environs = ['*']
    executable = 'echo bananas'

    @sanity_function
    def assert_bananas(self):
        return sn.assert_found(r'bananas', self.stdout)

    @run_after('setup')
    def hh(self):
        print('bananas')

@rfm.simple_test
class Derived(Base):
    def hh(self):
        print('not bananas')

Running this gives:

[==========] Running 2 check(s)
[==========] Started on Thu Jul  1 13:06:40 2021 

[----------] started processing Base (Base)
[ RUN      ] Base on dom:login using PrgEnv-gnu
bananas
[----------] finished processing Base (Base)

[----------] started processing Derived (Derived)
[ RUN      ] Derived on dom:login using PrgEnv-gnu
not bananas
[----------] finished processing Derived (Derived)

[----------] waiting for spawned checks to finish
[       OK ] (1/2) Base on dom:login using PrgEnv-gnu [compile: 0.006s run: 1.239s total: 1.266s]
[       OK ] (2/2) Derived on dom:login using PrgEnv-gnu [compile: 0.014s run: 1.316s total: 1.349s]
[----------] all spawned checks have finished

[  PASSED  ] Ran 2/2 test case(s) from 2 check(s) (0 failure(s), 0 skipped)
[==========] Finished on Thu Jul  1 13:06:42 2021 

The method Derived.hh still gets attached to the post_setup stage, despite that this method has not even been decorated with any run_before or run _after decorators.

[victorusu]

@jjotero, I think we need a unit test for this PR.
This behavior should be guaranteed to not break.
What do you think about adding your example as a unit test?
And also another unit test that re-attaches the hook?

[vkarak]

I have some minor comments + I agree with @victorusu to provide a unit test.

[codecov-commenter]

Codecov Report

Merging #2048 (148b16c) into master (2b2db4e) will decrease coverage by 0.01%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #2048      +/-   ##
==========================================
- Coverage   87.53%   87.51%   -0.02%     
==========================================
  Files          50       50              
  Lines        8889     8891       +2     
==========================================
  Hits         7781     7781              
- Misses       1108     1110       +2     

Impacted Files	Coverage Δ
reframe/core/hooks.py	90.90% <100.00%> (-1.88%)	⬇️
reframe/core/meta.py	98.96% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2b2db4e...148b16c. Read the comment docs.

[vkarak]

Lgtm, just a very minor comment.

[victorusu]

lgtm