Blockwise setindex in vcat etc.

[joschmitt]

This way we don't have to access single matrix entries for the flint types in Nemo.

Also removes vcat(::MatElem, ::MatElem) and hcat(::MatElem, ::MatElem) in favour of vcat(::MatrixElem...) etc. (assuming CI agrees).

[codecov]

Codecov Report

Attention: Patch coverage is 71.42857% with 4 lines in your changes are missing coverage. Please review.

Project coverage is 86.93%. Comparing base (e2b898a) to head (76d56a8).

Files	Patch %	Lines
src/Matrix.jl	71.42%	4 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1662      +/-   ##
==========================================
+ Coverage   86.88%   86.93%   +0.04%     
==========================================
  Files         116      116              
  Lines       29652    29624      -28     
==========================================
- Hits        25764    25753      -11     
+ Misses       3888     3871      -17     

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

[joschmitt]

So probably I can't remove the vcat(::MatElem, ::MatElem) version because it is used in OSCAR for the wrong reasons (see oscar-system/Oscar.jl#3580) and even if I fix that later versions of OSCAR would be broken with AbstractAlgebra, correct?

[lgoettgens]

So probably I can't remove the vcat(::MatElem, ::MatElem) version because it is used in OSCAR for the wrong reasons (see oscar-system/Oscar.jl#3580) and even if I fix that later versions of OSCAR would be broken with AbstractAlgebra, correct?

I would consider the current behavior here a bug (coercing the second matrix into the ring of the first), so this bug can and should be fixed. The fix in Oscar (oscar-system/Oscar.jl#3580) should get backported to 1.0.
And we will have a "fake breaking release" in the near future anyway (due to printing), so we could include this removal there to ease the transition.

And there seems to be some issue with Hecke here.

[thofma]

Do we need a new label? "fake breaking"?

[joschmitt]

I removed the commit that made this pull request breaking (and will open another pull request soon). This pull request should now be neither breaking nor fake breaking.