feat: improve performance on large SQL/graphs

[rubytobi]

As discussed in #679, I've done a first draft of minimal changes to speed up the column level lineage identification.

I think we should keep networkx - it's more powerful and flexible - and just switch to rustworkx where needed for performance reason.

Keen to hear your thoughts!

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.55%. Comparing base (10b7445) to head (2db8769).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #685   +/-   ##
=======================================
  Coverage   99.55%   99.55%           
=======================================
  Files          41       41           
  Lines        2244     2253    +9     
=======================================
+ Hits         2234     2243    +9     
  Misses         10       10           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[reata]

Can you write a performance test to show that rustworkx can handle large graph within certain time compared to networkx?

Also if we're converting networkx graph to rustworkx, then I don't think adding a graph property is needed. You can directly make a copy from self._sql_holder.graph.