Inline generic_matmatmul! branch in strided triangular matmul

[jishnub]

For combinations of strided matrices and strided triangular matrices, we would end up taking the methods defined in LinearAlgebra, so we may avoid the constant-propagation and hardcode the _generic_matmatmul! call. This improves TTFX, as the no-op but expensive-to-compile wrap call is elided.

julia> using LinearAlgebra

julia> A = zeros(4,4);

julia> @time A * UpperTriangular(A);
  0.458913 seconds (1.22 M allocations: 59.769 MiB, 51.63% gc time, 97.84% compilation time: 4% of which was recompilation) # master
  0.077198 seconds (174.52 k allocations: 8.683 MiB, 92.75% compilation time) # this PR

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.00%. Comparing base (0a253be) to head (331d7e2).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1262   +/-   ##
=======================================
  Coverage   92.00%   92.00%           
=======================================
  Files          34       34           
  Lines       15483    15485    +2     
=======================================
+ Hits        14245    14247    +2     
  Misses       1238     1238           

☔ 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.