Optimizations of L1 & L2 distance functors and miscellaneous improvements

[mlourakis]

Revised the unrolled loops in L1 & L2 distance calculations (L[12]_Adaptor):

• eliminated data dependency on d++
• added separate control paths depending on worst_dist
• unrolled processing of trailing elements with a fall-through switch

Additionally:

• added inline, constexpr, noexcept, __restrict specifiers
• added unit tests for L1
• other minor changes

[codecov]

Codecov Report

❌ Patch coverage is 72.72727% with 21 lines in your changes missing coverage. Please review.
✅ Project coverage is 84.25%. Comparing base (d8bfec8) to head (79c20b1).
⚠️ Report is 5 commits behind head on master.

Files with missing lines	Patch %	Lines
include/nanoflann.hpp	72.72%	21 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #277      +/-   ##
==========================================
- Coverage   86.20%   84.25%   -1.96%     
==========================================
  Files           1        1              
  Lines         667      705      +38     
  Branches      125      130       +5     
==========================================
+ Hits          575      594      +19     
- Misses         92      111      +19     

Files with missing lines	Coverage Δ
include/nanoflann.hpp	84.25% <72.72%> (-1.96%)	⬇️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[jlblancoc]

Awesome, thanks a lot! 👏

Please, see my minor comments above.

[jlblancoc]

@mlourakis LGTM, thanks again for the contribution!

On alignment: totally agree. Your comments are perfect as is now.
We could add macros to auto-detect the enabled compiler flags (SSE*, AVX*, etc.) and architecture and define the alignment accordingly... my only fear is it might become ABI incompatible (-> crashes) if at some point nanoflann.hpp is included from two translation units with different compiler flags (same problem happened with Eigen3 in the past... 😢 ).

[mlourakis]

Fully agree about autodetecting the HW details, that's one of the reasons I was reluctant to attempt it. But someone interested in getting the best possible performance, might choose to tweak alignment on his own machine.

Thanks for your work on nanoflann!