Fix undefined tol and remove unused variable in QR and LQ pullback

[leburgel]

The gauge dependency checks in qr_pullback! and lq_pullback! threw an error because the check tol was defined in a let block. Also, the ΔR11 variable was unused in both cases, so I removed it.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 81.94%. Comparing base (7b480fc) to head (9789de2).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #127      +/-   ##
==========================================
- Coverage   81.95%   81.94%   -0.01%     
==========================================
  Files          42       42              
  Lines        5680     5678       -2     
==========================================
- Hits         4655     4653       -2     
  Misses       1025     1025              

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

[lkdvos]

I don't think this matters for performance all that much, but in principle you could check if this leads to a "boxed" variable inside of the closure used by findlast. The let block avoids this which could lead to more type stability, but it probably does not make a huge difference. You could check this by descending into it via Cthulhu.jl, which should show some warnings.

[leburgel]

I could check, but if it's anyway safer I could also just put back the let block and redefine the tolerance later when it's used for the check (potentially again in a scoped block, though I don't know if it makes any difference for that specific use). I definitely wasn't thinking about this, just wanted it to run.

In any case, the way I changed it here is exactly the same as what was already happening in the SVD pullback. So if this is something we should be careful with, I can also change it there.

[Jutho]

Thanks.