FEAT: Impl collective communication interface

[ChengjieLi28]

Related #22

Tested on xorbits.

In the case of worker number 5 for tpch 100, broadcast is much more slower than copy_to.
Maybe when cluster scale is very large, collective communication may speed up. But I do not know now.
See issue: https://stackoverflow.com/questions/26558189/mpi-broadcast-very-slow

Check code requirements

• tests added / passed (if needed)
• Ensure all linting tests pass

[codecov]

Codecov Report

Merging #46 (cac0e28) into main (c3ac228) will increase coverage by 0.34%.
The diff coverage is 95.51%.

@@            Coverage Diff             @@
##             main      #46      +/-   ##
==========================================
+ Coverage   93.66%   94.00%   +0.34%     
==========================================
  Files          42       47       +5     
  Lines        3361     3671     +310     
  Branches      672      708      +36     
==========================================
+ Hits         3148     3451     +303     
- Misses        141      145       +4     
- Partials       72       75       +3     

Flag	Coverage Δ
unittests	93.87% <95.51%> (+0.35%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files Changed	Coverage Δ
python/xoscar/collective/utils.py	63.63% <63.63%> (ø)
python/xoscar/collective/core.py	95.57% <95.57%> (ø)
python/xoscar/collective/process_group.py	96.71% <96.71%> (ø)
python/xoscar/backends/communication/ucx.py	94.61% <100.00%> (+2.27%)	⬆️
python/xoscar/collective/__init__.py	100.00% <100.00%> (ø)
python/xoscar/collective/common.py	100.00% <100.00%> (ø)

... and 2 files with indirect coverage changes

[aresnow1]

LGTM