Backport PR #47393 on branch 1.4.x (CI/TST: Don't require length for construct_1d_arraylike_from_scalar cast to float64)

[meeseeksmachine]

Backport PR #47393: CI/TST: Don't require length for construct_1d_arraylike_from_scalar cast to float64

[simonjayhawkins]

urgh. Numpy Dev still fails

[simonjayhawkins]

not yet got down to why this is not failing in main. made the code change anyway for now to see if tests pass.

[simonjayhawkins]

not yet got down to why this is not failing in main. made the code change anyway for now to see if tests pass.

@mroeschke Numpy Dev is green. waiting on the others. OK to merge or allow time to investigate further? (and fix for 1.4.4 instead). I don't think this is urgent.

[mroeschke]

not yet got down to why this is not failing in main. made the code change anyway for now to see if tests pass.

@mroeschke Numpy Dev is green. waiting on the others. OK to merge or allow time to investigate further? (and fix for 1.4.4 instead). I don't think this is urgent.

Agreed this is not urgent, but the CI looks "green" (Windows failures appear to the memory errors and the ASV failure is a conda connection error"). So can go in 1.4.3 or 1.4.4

[simonjayhawkins]

OK let's do for 1.4.3. Since this is numpy-dev related it doesn't affect wheel builds. but the code changes will get exercised by users and if any issues identified could then be resolved in 1.4.4.