The fast-path QImage creation methods residing within ImageItem can be factored out to be free-standing functions.
This has the following benefits:

1. Makes the routine stateless, which ensures that it has no side-effects from other methods.
2. Allows the functionality to be used outside of ImageItem

Some behavior changes:

1. replaces the use of nextafter with the snippet in Fix UnderflowError in _try_rescale_float when maxVal == 0. #2881 (comment)
2. doesn't cache the effective lut.
   • Caching the effective lut makes the code stateful.
   • The effective lut computation is relatively small. While the overhead of recomputing the effective lut may be significant for very small images, displaying such small images should be fast enough already.
     • The largest computation that this caching would have saved is for the case "uint16 image and lut with more than 256 entries". This results in an effective lut size of 128KB. However, since luts are usually created with 256 entries, this case would be seldom encountered.
3. the test of ImageFormat no longer needs to instantiate an ImageItem object for each test. This could be problematic if the CI was flakey to begin with.