Determine eltype and ndims type parameter positions automatically

[mtfishman]

This brings back the functionality removed in #26 but with a more robust strategy.

This uses the fact that when an AbstractArray subtype with unconventional type parameter positions for eltype and ndims is defined, the information of where those type parameters are defined can be determined by matching the names of the TypeVars of the type parameters to the same name in the AbstractArray type output by InteractiveUtils.supertypes, for example:

julia> struct MyBoolArray{X,Y,Z,B} <: AbstractArray{Bool,B} end

julia> supertypes(MyBoolArray)
(MyBoolArray, AbstractArray{Bool}, Any)

julia> Base.unwrap_unionall(supertypes(MyBoolArray)[1]).parameters
svec(X, Y, Z, B)

julia> Base.unwrap_unionall(supertypes(MyBoolArray)[2]).parameters
svec(Bool, B)

So you can see that we can determine that the 4th parameter of MyBoolArray must be the ndims type parameter, since it has a name B matching the ndims parameter of its AbstractArray supertype.

In principle this can work with non-AbstractArray types (using the internal function TypeParameterAccessors.position_from_supertype) but for now that is opt-in and an internal feature.

[codecov]

Codecov Report

Attention: Patch coverage is 79.16667% with 5 lines in your changes missing coverage. Please review.

Project coverage is 83.41%. Comparing base (24b3add) to head (fc8c8df).
Report is 3 commits behind head on main.

Files with missing lines	Patch %	Lines
src/base/abstractarray.jl	40.00%	3 Missing ⚠️
src/type_parameters.jl	89.47%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #31      +/-   ##
==========================================
+ Coverage   83.08%   83.41%   +0.32%     
==========================================
  Files          14       14              
  Lines         201      217      +16     
==========================================
+ Hits          167      181      +14     
- Misses         34       36       +2     

Flag	Coverage Δ
docs	43.92% <30.43%> (-1.53%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[lkdvos]

Thanks for taking the time to clarify, it's definitely easier to follow for me like this 😄

[mtfishman]

Thanks for taking the time to clarify, it's definitely easier to follow for me like this 😄

I was also getting confused by the previous code I wrote as I was rewriting it so that wasn't a good sign!