Persist original ARM API version for use by generic controller

[theunrepentantgeek]

What this PR does / why we need it:

Adds OriginalVersion() function to API Spec types to return the actual ARM API version requested by users.
Adds OriginalVersion property to Storage Spec types to store that version for later use.
Adds OriginalGVK() function to Storage Resource types to return the GVK of the original version.
Removes StorageTypeFactory in favour of separate pipeline sections

Also includes

• Unit tests for OriginalVersion() function generation
• Unit tests for OriginalGVK() function generation
• Fix bug in prerequisite checks for pipeline stages
• Fix bug in ErroredType where WriteDebugDescription() would crash if there was no wrapped type

How does this PR make you feel:

If applicable:

• this PR contains tests

[codecov-commenter]

Codecov Report

Merging #1606 (8ebde28) into master (1e458b2) will increase coverage by 1.89%.
The diff coverage is 80.25%.

❗ Current head 8ebde28 differs from pull request most recent head 393d2a7. Consider uploading reports for the commit 393d2a7 to get more accurate results

@@            Coverage Diff             @@
##           master    #1606      +/-   ##
==========================================
+ Coverage   62.92%   64.82%   +1.89%     
==========================================
  Files         187      200      +13     
  Lines       12359    12542     +183     
==========================================
+ Hits         7777     8130     +353     
+ Misses       3897     3702     -195     
- Partials      685      710      +25     

Impacted Files	Coverage Δ
hack/generator/pkg/astmodel/resource_type.go	79.02% <0.00%> (+5.44%)	⬆️
hack/generator/pkg/astmodel/type_walker.go	67.53% <ø> (ø)
hack/generator/pkg/astmodel/types.go	59.37% <ø> (+2.08%)	⬆️
...enerator/pkg/codegen/storage/property_converter.go	57.81% <ø> (+57.81%)	⬆️
...g/testcases/object_type_serialization_test_case.go	2.06% <0.00%> (-0.04%)	⬇️
hack/generator/pkg/astmodel/errored_type.go	33.33% <50.00%> (ø)
...n/pipeline/inject_property_assignment_functions.go	66.66% <66.66%> (ø)
...ck/generator/pkg/codegen/storage/type_converter.go	69.81% <66.66%> (+69.81%)	⬆️
hack/generator/pkg/astmodel/function_container.go	71.42% <71.42%> (ø)
hack/generator/pkg/astmodel/property_container.go	71.42% <71.42%> (ø)
... and 48 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1e458b2...393d2a7. Read the comment docs.

[matthchr]

Not done yet but posting what I have

[babbageclunk]

This looks great, although I think the change you discussed of breaking StorageTypeFactory into a set of pipeline stages will be nicer.

I've majorly restructured the PR, so this approval is stale.

[matthchr]

LGTM - Left a bunch of nits and one comment about golden test file structure.

[matthchr]

extra mega minor:

Suggested change

	//TODO: For now only used for ARM
	// TODO: For now only used for ARM

[matthchr]

I hate that go doesn't have first class enum support, and the solution of consts kinda works but if the consts don't have the enum name in them it's sometimes hard to pick out which values belong to which type.

I know in some places we do something like OriginalVersionKindReadProperty (which is sorta how you'd do it in C, I think?), but we don't have that consistently and it is awkwardly long.

Does this bug other folks too? What do people prefer?

[theunrepentantgeek]

I used very simple names here to optimize the API when consuming; adding a unique suffix would make that clunky.

Having first class enum support would be preferable - but I guess Go is designed to push people towards many simpler packages.