cast raster math inputs to 32-bit floating-point by default to avoid unexpected underflows/overflows

[Andreas Janz]

Originally requested by @Agustin Lobo in #897.

Add a checkbox “Use 32-bit floating-point inputs” that is checked by default.

[Agustin Lobo]

I would suggest “Use 32-bit floating-point operations”, and the inputs would be cast if needed.

Also, the output should let the user select file name, file format, data format (i.e., FLT32, UINT16 etc) and band order (BSQ/BIL/BIP)

[Agustin Lobo]

Any expected due date for this improvement?

[Andreas Janz]

Planned for v3.10, but it is easy to implement, so I can sneak it in soon.

[Andreas Janz]

resolves #908

→ <<cset 193afc604fc4>>