Feature: Allow configuration of custom enum markers for `maxlength` linter

[everettraven]

In OpenShift, we have some custom markers that set enum values for a field and this results in the maxlength linter stating that a field/type alias should have a maximum length when using this custom marker instead of the standard kubebuilder:validation:Enum marker.

While this particular case is OpenShift-specific, I think it is reasonable to make a generic way to extend this detection logic as there may be other vendors and/or projects that use their own custom markers for CRD generation.

[everettraven]

@JoelSpeed What are your thoughts here?

[JoelSpeed]

In this particular case, we actually can then have multiple declarations of the MaxLength which are applied in different scenarios. Does this linter need to be able to determine the difference between the multiple values, or does it just check for the presence of a single value?

I think in most of the rules we have, adding the ability to customise the set of markers that the rule is looking for probably makes sense

[everettraven]

Circling back around to this.

In this particular case, we actually can then have multiple declarations of the MaxLength which are applied in different scenarios. Does this linter need to be able to determine the difference between the multiple values, or does it just check for the presence of a single value?

I suspect you are referring to the fact that we may have multiple feature-gated validations that enforce different MaxLength values depending on the feature-gate that is active?
AFAIK, the maxlength linter just checks to see if there should be a MaxLength provided so I don't think it quite captures the nuance of the situation here.

I'll have to do a bit more thinking on this, but I'm realizing that the feature-gate aspect of this throws some of the expectations we have off kilter a bit. To properly handle the feature-gate aware validations, we would need to be able to iterate through the different validation "sets" for each feature-gate to ensure they make sense to the linter. This is something that would likely need to be a more nuanced translation layer that may be better suited to be handled on a case-by-case basis rather than trying to shoehorn something generic here.

[JoelSpeed]

I think in this one, I'm thinking along the lines of, today, where it currently just checks, "is there a XYZ marker", then that's fine and works. But we know from our own experience, that there's a possibility that we need "is there an XYZ marker with a keyed value". So, I'm trying to work out, does this linter care about knowing the value of the maximum? If it does, then when we allow custom markers, we would also need to specify if we expect the value to be under a certain key, or, whether we expect it to just be present, and presence is sufficient