Fix local_size() in some patterns and improve short-cut detection

[devreal]

Fixes #653

[codecov]

Codecov Report

Merging #655 into development will increase coverage by 0.07%.
The diff coverage is 97.5%.

@@               Coverage Diff               @@
##           development     #655      +/-   ##
===============================================
+ Coverage        85.06%   85.13%   +0.07%     
===============================================
  Files              336      336              
  Lines            24954    24984      +30     
  Branches         11803    11358     -445     
===============================================
+ Hits             21226    21271      +45     
+ Misses            3693     3691       -2     
+ Partials            35       22      -13

Impacted Files	Coverage Δ
dash/include/dash/pattern/BlockPattern.h	99.15% <100%> (ø)	⬆️
dash/test/container/MatrixTest.cc	99.35% <100%> (+0.01%)	⬆️
dash/include/dash/pattern/TilePattern.h	95.8% <100%> (+0.26%)	⬆️
dash/include/dash/pattern/BlockPattern1D.h	99.49% <100%> (ø)	⬆️
dash/include/dash/pattern/SeqTilePattern.h	92.77% <100%> (+0.38%)	⬆️
dash/include/dash/pattern/TilePattern1D.h	89.87% <85.71%> (-0.52%)	⬇️
...h/include/dash/iterator/internal/ContiguousRange.h	98.76% <0%> (+1.23%)	⬆️
dash/include/dash/Team.h	87.62% <0%> (+3.09%)	⬆️
... and 1 more

[bertwesarg]

Solves the infinite loop in #653, but not the example fully.

[devreal]

Please review

[bertwesarg]

I'll pass. Cannot add something to this. If it works, than got ahead.