Remove glibc constraint from aarch64 packages

[charlesbluca]

Addresses conda-forge/dask-sql-feedstock#66 by reordering imports to avoid the aarch64 glibc TLS allocation bug and removing the glibc from the conda / pip packages.

[codecov-commenter]

Codecov Report

Merging #1201 (619a068) into main (c6b2ce1) will increase coverage by 0.11%.
The diff coverage is 100.00%.

❗ Your organization is not using the GitHub App Integration. As a result you may experience degraded service beginning May 15th. Please install the Github App Integration for your organization. Read more.

@@            Coverage Diff             @@
##             main    #1201      +/-   ##
==========================================
+ Coverage   82.12%   82.23%   +0.11%     
==========================================
  Files          78       78              
  Lines        4571     4572       +1     
  Branches      849      849              
==========================================
+ Hits         3754     3760       +6     
+ Misses        634      625       -9     
- Partials      183      187       +4     

Impacted Files	Coverage Δ
dask_sql/__init__.py	100.00% <100.00%> (ø)

... and 1 file with indirect coverage changes

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

[charlesbluca]

Tested the packages/wheels on linux-aarch64 with glibc<2.32, can confirm that this allows us to import successfully in an environment with scikit-learn installed, will note that the following still doesn't work:

import sklearn   # does some TLS allocations
import dask_sql  # now there isn't any space for the dask_planner.rust allocations

But seems like the preferable solution here from the conda-forge perspective is to broadcast the presence of the bug and recommend that users pursue workarounds on their own rather than outright blocking installation, so this should be good to merge

[jakirkham]

Awesome work! 👏

Thanks Charles & Ayush 🙏