Store global node IDs for partitioned graphs

[kaixuanliu]

@rusty1s For distributed pyg training development, we prefer to use global node ids to represent partitioned graph store. It would be convenient for us to neighbor sample the partitioned graph if we use global ids for rows/cols.

[kaixuanliu]

@ZhengHongming888

[codecov]

Codecov Report

Merging #7910 (5c96900) into master (deff5a4) will decrease coverage by 0.80%.
Report is 2 commits behind head on master.
The diff coverage is 100.00%.

❗ Current head 5c96900 differs from pull request most recent head e18b7b9. Consider uploading reports for the commit e18b7b9 to get more accurate results

@@            Coverage Diff             @@
##           master    #7910      +/-   ##
==========================================
- Coverage   90.16%   89.36%   -0.80%     
==========================================
  Files         457      457              
  Lines       26821    26826       +5     
==========================================
- Hits        24182    23974     -208     
- Misses       2639     2852     +213     

Files Changed	Coverage Δ
torch_geometric/distributed/partition.py	99.15% <100.00%> (+0.06%)	⬆️

... and 33 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more