Fix vector_norm wrong results when order is inf

[nikifaets]

Close #10480

[nikifaets]

Kudos to @suvadityamuk for deep diving into the bug, finding the root causes and proposing a solution.

[nikifaets]

I added specific test examples that cover the cases when ord is inf or -inf.

Currently, the ivy hypothesis_helpers.number_helpers.floats method does not support the allow_infinity argument that comes with the Hypothesis implementation of floats. Unless I'm missing something, this prevents us from easily sampling infs as the ord value in the test cases by. At least not by modifying the @handle_test annotation.

We can use hypothesis.strategies.floats instead of helpers.floats and use the allow_infinity argument for the ord values but this does not guarantee that an inf would be drawn every time the tests are run.

I'm open to suggestions on how to include inf examples in the generic testing method (and be sure that there are infs every time the test is run).

[iamjameskeane]

The logic seems sound to me but might be more appropriate for @suvadityamuk to give this a check over before merging :)

[suvadityamuk]

LGTM. Checked tests, they seem to pass.

[suvadityamuk]

@jkeane508 Please feel free to merge!

@nikifaets Thank you for your contribution!