Coloring and decompression for structured matrices

[gdalle]

Fixes #64 and fixes #65 by implementing structured coloring & decompression for various structured matrix types:

Package	Type	Coloring	Decomp.	Test perf
LinearAlgebra	Diagonal	✅	✅	✅
LinearAlgebra	Bidiagonal	✅	✅	✅
LinearAlgebra	Tridiagonal	✅	✅	✅
LinearAlgebra	SymTridiagonal
BandedMatrices	BandedMatrix	✅	✅
BlockBandedMatrices	BlockBandedMatrix	✅
BlockBandedMatrices	BandedBlockBandedMatrix	✅

General philosophy

Keep the same result type (ColumnColoringResult / RowColoringResult).

• Upside: if the user tries to decompress into something else, like a SparseMatrixCSC, it will still be very fast
• Downside: a few needless allocations after coloring, but unlike decompression they are a one-time cost.

Remaning issues

For SymTridiagonal I'm not sure what to do since it is not mutable.

For decompression of BlockBandedMatrices, I still need to figure out JuliaArrays/BlockArrays.jl#418. But it already works because the fallback algorithm relies on the nonzeros of the sparse encoding, so it does not try to access illegal indices outside of the bands.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 100.00%. Comparing base (1cb3884) to head (df53bde).

Additional details and impacted files

@@            Coverage Diff             @@
##              main      #132    +/-   ##
==========================================
  Coverage   100.00%   100.00%            
==========================================
  Files           11        14     +3     
  Lines          871      1022   +151     
==========================================
+ Hits           871      1022   +151     

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

[gdalle]

Superseded by #139 (I screwed up with a git rebase while splitting out the tests)