Replace arrayCoordinateSystem with explanation on how to express dimensionless transforms in pixel coordinates

[jo-mueller]

Fixes ome/ngff#438
Addresses comment 4 on RFC5.

Title. This PR replaces the definition of implicit arrayCoordinateSystems with examples and explanatory text on how to properly express transforms that are done in pixel space, not unit space.

Follow up PR to #103 and #90

cc @bogovicj @btbest @will-moore @clbarnes @dstansby

[github-actions]

Automated Review URLs

• Readthedocs

[btbest]

Perfect from my point of view! 🚀

[bogovicj]

I see this PR's scope as being pretty small, it does what it says and is clear (modulo the discussion below). So I approve these changes.

There are comments here about definitions and terminology that deserve further discussion. @d-v-b opened this issue ome/ngff#492
and I suggest we move the discussion there.

It's true that the phrasing of the examples and text edited here will change as a result of those decisions we'll have about terminology, but they'll affect more than what is written here, so I'd suggest those larger changes be a separate PR.

[will-moore]

I never really felt the need for Array coordinate systems, so it almost feels a bit verbose to describe how to work without them. But no objections if it's useful, Thanks

[jo-mueller]

so it almost feels a bit verbose to describe how to work without them. But no objections if it's useful

Yeah...maybe we'll need a section for "How to express this or that data structure" to be broken out of the spec at some point 🤔