Replace multiprocessing with joblib for parallelization and change n_cores default to 1

[tsalo]

Closes None.

Changes proposed in this pull request:

• Use joblib instead of multiprocessing for parallelization.
• Drop if/else statements based on number of cores. Just use joblib regardless.
  • This may slow things down when n_cores is 1, but I want to be sure before using more complicated code.
• Change default number of cores to 1 in all tools that have n_cores as a parameter.
• Hopefully address issues with unclosed memmaps (see Bus errors/"too many open files" errors in jobs with large numbers of Estimators using temporary files #631).

To do:

• Drop tmpdir change now in Use tmpdir for memmap files instead of the NiMARE data directory #599.
• Do some profiling.
  • At least on MKDADensity, without correction, joblib and multiprocessing perform the same.
• Change default n_cores to 1 throughout library.

[codecov]

Codecov Report

Merging #597 (fedd3a8) into main (c35ca9b) will increase coverage by 0.00%.
The diff coverage is 93.45%.

@@           Coverage Diff           @@
##             main     #597   +/-   ##
=======================================
  Coverage   87.89%   87.89%           
=======================================
  Files          40       40           
  Lines        4320     4337   +17     
=======================================
+ Hits         3797     3812   +15     
- Misses        523      525    +2     

Impacted Files	Coverage Δ
nimare/workflows/ale.py	92.42% <ø> (ø)
nimare/workflows/macm.py	25.71% <ø> (ø)
nimare/workflows/scale.py	80.55% <ø> (ø)
nimare/meta/cbma/ale.py	92.74% <84.21%> (-1.41%)	⬇️
nimare/utils.py	95.41% <95.45%> (-0.06%)	⬇️
nimare/base.py	88.18% <100.00%> (ø)
nimare/diagnostics.py	95.29% <100.00%> (-0.11%)	⬇️
nimare/meta/cbma/base.py	94.69% <100.00%> (+0.55%)	⬆️
nimare/meta/cbma/mkda.py	96.31% <100.00%> (ø)
nimare/meta/ibma.py	100.00% <100.00%> (ø)
... and 1 more

---

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 c35ca9b...fedd3a8. Read the comment docs.

[tsalo]

Joblib can work with memmaps: https://joblib.readthedocs.io/en/latest/parallel.html#working-with-numerical-data-in-shared-memory-memmapping

Also, there's an example: https://joblib.readthedocs.io/en/latest/auto_examples/parallel_memmap.html#sphx-glr-auto-examples-parallel-memmap-py