ENH: Use pure SQL for return_type = 'id' queries

[adelavega]

An alternative to #948 and #942

Instead of relying on looping over entities, allow the core logic of .get to return a subquery, rather than list of files.

This subquery can then be used to find the unique entities of a given target for a list files, in pure SQL.

This can a bit faster (22ms vs 6-7 ms on my test dataset w/ ~80 subs), which could make a difference on very large datasets.

[adelavega]

I would say we hold off on this until we have developed benchmarks that justify the potential added complexity

[codecov]

Codecov Report

Patch coverage: 82.60% and project coverage change: +0.02% 🎉

Comparison is base (ae1bb00) 83.46% compared to head (0079807) 83.48%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #949      +/-   ##
==========================================
+ Coverage   83.46%   83.48%   +0.02%     
==========================================
  Files          38       38              
  Lines        4305     4311       +6     
  Branches     1096     1098       +2     
==========================================
+ Hits         3593     3599       +6     
  Misses        514      514              
  Partials      198      198              

Files Changed	Coverage Δ
bids/layout/layout.py	88.60% <82.60%> (+0.13%)	⬆️

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