Update imgData endpoint to return more metadata about the resolutions

[sbesson]

Fixes #584

The response to the webgateway/imgData endpoint uses ImageWrapper.getZoomLevelScaling which reduces the resolution description as a one dimensional array of scalar zoom levels.
This reduction suffers from limitations both due to rounding errors as well as resolution levels where sizeX and sizeY might have different zoom level scaling.
This commit marshals the sizeX/sizeY pair returned from PixelBuffer.getResolutionDescriptions into a new dictionary for all resolutions and returns it as an additional resolutions field in the JSON response for multi-resolution images.

The proposed value of the resolutions attribute mirrors zoomLevelScaling by using a dictionary where each key is the resolution index and only being populated for images with more than one resolution. As a possible discussion point, resolutions could be populated for all images independently of the number of resolution level.

/cc @melissalinkert @kkoz

[sbesson]

Once the initial draft has been validated by the nightly CI builds to confirm there is no regression, the plan is to extend the https://github.com/ome/openmicroscopy/blob/develop/components/tools/OmeroWeb/test/integration/test_marshal.py integration. test to cover the multi-resolution scenario in addition to the simple small multi-dimensional image.

[knabar]

Will this change the behavior of zoomLevelScaling? Currently for images with a single level, zoomLevelScaling is None, whereas now it will be a dictionary with one entry.

Clients that check for the existence of a zoomLevelScaling value may have to be adjusted.

Could we alternatively keep the current zoomLevelScaling and populate a new entry instead?

[sbesson]

No, the intent of this is to be 100% backwards-compatible although zoomLevelScaling is effectively made redundant by the resolutions attribute as the client could use its value to calculate the scaling.

My reading of the current implementation is that for single-resolution images, the tiles boolean will be False and the zoomLevelScaling population will be skipped altogether as per

omero-web/omeroweb/webgateway/marshal.py

Lines 183 to 186 in 1c82ac1

	levels = image._re.getResolutionLevels()
	tiles = levels > 1
	rv["tiles"] = tiles
	if tiles:

. This is also what is being tested in https://github.com/ome/openmicroscopy/blob/0763546fbc345dbc1bbcf1995bd9d631f5ba1567/components/tools/OmeroWeb/test/integration/test_marshal.py#L52-L55

[will-moore]

The test has assert 'zoomLevelScaling' not in img_data but with this PR, img_data will always contain zoomLevelScaling.

I understand we don't want to call _re.getResolutionDescriptions() more than once, but if there's only a single item then don't set zoomLevelScaling

[knabar]

My reading of the current implementation is that for single-resolution images, the tiles boolean will be False and the zoomLevelScaling population will be skipped altogether as per

After another reading I agree, the presence of zoomLevelScaling should not change. If tests pass, 👍 from me.

[jburel]

Tests are green

[will-moore]

LGTM 👍

[mabruce]

Changes look good to me.