Fix dt_camctl_camera_set_property_float brokenness

[victoryforce]

The issue was caused by undetected copy-paste errors. dt_camctl_camera_set_property_float() looks like copy of the _int variant, in which not all the necessary changes have been made. Allocation was used size of wrong struct, fortunately their sizes was the same (as int and float both is 4 bytes on most modern architectures).

The direct cause of the buggy behavior is the following assignment:

job->type = _JOB_TYPE_SET_PROPERTY_INT;

This type is used to select an option in the switch in the _camera_process_job() function. Because of this erroneous type, an execution path for an integer type will be chosen there, in which case the value from job->value will eventually be assigned to the int variable.

As a result, the fractional part will be discarded (not mathematical rounded, BTW), and the number will be stored as an integer. This means that the fractional properties of the camera will be significantly distorted.

[victoryforce]

@TurboGit - As for the release notes, I don't use camera tethering myself. Also, I'm currently working on Windows, where I don't want to bother with tethering support tricks. So I can't see how exactly this bug manifests itself in the UI yet (I didn't encounter it while working with darktable, I found it while reading the code). If someone checks and reports it, I could add this wording to the release notes...

[TurboGit]

I would say no release notes needed for this, never complained and this is certainly not a widely used feature. Thanks!