Bugfix: Correct size for non-CUDA device copies

[mlange05]

A small bug-fix that was discovered with IFS regression testing. The incorrect size of the copy buffer was triggering cuMemcpyHtoDAsync errors, which are fixed with this.

Many thanks to @awnawab for debugging help and @pmarguinaud for providing the test.

[awnawab]

Hi @jameshawkes. We seem to be having some trouble with the fussy-duck and would really appreciate your help. As you can see, we're getting an error saying the CLA has not been signed. However when we click on it we get the following message:

Error
There is no CLA to sign for ecmwf-ifs/field_api

(Resource protected by organization SAML enforcement. You must grant your Personal Access token access to this organization.)

What's odd about this error is that both @mlange05 and I have granted sso authorization to ecmwf-ifs. Have you seen something like this before?

[FussyDuck]

All committers have signed the CLA.

[FussyDuck]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

:white_check_mark: pmarguinaud
:x: mlange05
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[FussyDuck]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

:white_check_mark: pmarguinaud
:x: mlange05
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[jameshawkes]

Should be good now, just @mlange05 that needs to sign. Sorry about that!

[awnawab]

Thanks a lot @jameshawkes for sorting this out 🙏