BUG: calculate_distance_matrix computes all pairs

[bifbof]

fixes #592
One of the added tests has a very generous relative tolerance, this is due to #593.
I made this function probably quite a bit slower. I thought that correctness is more important and if nobody experienced this bug in 3 years maybe it isn't called that often either.
Tried to use pairwise distance for everything. That wasn't possible for the LineStrings as sklearn and scipy both enforce float as data type. Thus I copied the code out of sklearn and adapted it a bit.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (e98354d) 93.15% compared to head (061299e) 93.30%.

❗ Current head 061299e differs from pull request most recent head c1b4182. Consider uploading reports for the commit c1b4182 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #594      +/-   ##
==========================================
+ Coverage   93.15%   93.30%   +0.14%     
==========================================
  Files          33       33              
  Lines        2104     2076      -28     
  Branches      374      368       -6     
==========================================
- Hits         1960     1937      -23     
+ Misses        130      127       -3     
+ Partials       14       12       -2     

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

[hongyeehh]

Thanks for the fix! The current version is much cleaner and easier to maintain. I only have a minor comment relating to n_jobs docstring