Out-of-place triu/tril for Symmetric in each branch

[jishnub]

In tril/triu for Symmetric/Hermitian, we should use the out-of-place variants in each branch, unless we're sure that the array may be mutated. Currently, this is only for StridedArrays, in which case we may perform some of the operations in-place.

Fixes issues like

julia> S = @SMatrix [1 2; 3 4]
2×2 SMatrix{2, 2, Int64, 4} with indices SOneTo(2)×SOneTo(2):
 1  2
 3  4

julia> H = Hermitian(S)
2×2 Hermitian{Int64, SMatrix{2, 2, Int64, 4}} with indices SOneTo(2)×SOneTo(2):
 1  2
 2  4

julia> triu(H,-1)
ERROR: setindex!(::SMatrix{2, 2, Int64, 4}, value, ::Int) is not defined.
 Hint: Use `MArray` or `SizedArray` to create a mutable static array

After this PR, this operation would work:

julia> triu(H,-1)
2×2 MMatrix{2, 2, Int64, 4} with indices SOneTo(2)×SOneTo(2):
 1  2
 2  4

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 93.81%. Comparing base (39ee3af) to head (ae9206e).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1318   +/-   ##
=======================================
  Coverage   93.81%   93.81%           
=======================================
  Files          34       34           
  Lines       15715    15719    +4     
=======================================
+ Hits        14743    14747    +4     
  Misses        972      972           

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

🚀 New features to boost your workflow:

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

[dkarrasch]

Are there cases where the former doesn't do what we want?

[jishnub]

I can't think of any immediately. This was an overly conservative change that may not be necessary in practice.