[jdbc.read] Properly pre-compute column fetching thunk

[alexander-yakushev]

Thanks to your comment on #179 I finally understood what is the deal with column thinks and such. This PR moves computing the thunk to where it should be – outside of the returned lambda, so that the machinery is not re-run for every row. The benchmark from #185 gives this:

master (results are reused):
Time per call: 18.89 ms   Alloc per call: 17,484,278b
Time per call: 18.90 ms   Alloc per call: 17,483,998b
Time per call: 18.87 ms   Alloc per call: 17,483,817b

After PR:
Time per call: 9.83 ms   Alloc per call: 11,968,423b   Iterations: 209
Time per call: 9.99 ms   Alloc per call: 11,968,262b   Iterations: 201
Time per call: 9.76 ms   Alloc per call: 11,968,255b   Iterations: 209

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 83.57%. Comparing base (352d181) to head (5b14028).
Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #189      +/-   ##
==========================================
- Coverage   83.58%   83.57%   -0.01%     
==========================================
  Files          37       37              
  Lines        2497     2496       -1     
  Branches      212      212              
==========================================
- Hits         2087     2086       -1     
  Misses        198      198              
  Partials      212      212              

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

[camsaul]

Oh wow, that was 100% the way it was supposed to work in the first place. Thanks for catching this