Avoid construction in convert

[jishnub]

This follows the convert implementation for AbstractArrays and short-circuits if the input is already of the correct type. There's no actual construction required in this case.

Should fix the failure on master seen in #1554

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.04%. Comparing base (3a4fdad) to head (00898d1).
⚠️ Report is 2 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1556   +/-   ##
=======================================
  Coverage   94.04%   94.04%           
=======================================
  Files          35       35           
  Lines       15986    15990    +4     
=======================================
+ Hits        15034    15038    +4     
  Misses        952      952           

☔ 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]

Do you have any idea why this wasn't an issue before?

[jishnub]

The method was seemingly broken in #1212. It was working correctly on v1.12 before that PR was merged. I'm not entirely certain why this test didn't appear before. The Transpose seems to work even without the convert:

julia> B
4×4 Bidiagonal{Int64, UnitRange{Int64}}:
 1  1  ⋅  ⋅
 ⋅  2  2  ⋅
 ⋅  ⋅  3  3
 ⋅  ⋅  ⋅  4

julia> Transpose(B)
4×4 transpose(::Bidiagonal{Int64, UnitRange{Int64}}) with eltype Int64:
 1  ⋅  ⋅  ⋅
 1  2  ⋅  ⋅
 ⋅  2  3  ⋅
 ⋅  ⋅  3  4

julia> convert(typeof(B), B)
ERROR: MethodError: Cannot `convert` an object of type Vector{Int64} to an object of type UnitRange{Int64}

julia> VERSION
v"1.14.0-DEV.1818"

I can replicate the failure on using julia's commit 248531f9b6 but this works on e8947b0ecfc, so I wonder if this was just a regression on nightly.

Edit: Looks like JuliaLang/julia@3e744a7 was the fix