FIX: unintended conversion of float to numpy.float64

[juhi24]

Rationale

Fix #2472

Implications

I believe this is how the original authors intended it. I believe the only implication should be fixing the reported bug, i.e. avoiding unwanted type conversion.

[CLAassistant]

All committers have signed the CLA.

[greglucas]

This seems like an issue upstream with fetch_raster and their Java implementation not handling/converting numpy datatypes. This seems simple enough though that we can easily add it.

I'm not sure there is a great test we could add here, so maybe just a comment would be helpful explaining why we are doing this?

[juhi24]

The #2472 was fixed upstream. The fix proposed in this PR is not strictly necessary. It's now a guestion between little added robustness and simplicity. May the maintainers decide.

Fixed upstream, so this is unnecessary here now.

[greglucas]

Because this is fixed upstream, I do not think we should add it here anymore, that is adding an extra calculation that isn't needed. Thanks for following up and getting it fixed upstream!