vertex-ai: add :UPLOAD support for model resource

[BBBmau]

adding support of :UPLOAD for models resource. Once completed will be merged into base models PR: #12490

• :UPLOAD
  • explanation_spec_examples vertex-ai: add :UPLOAD explanation_spec_examples tests  #12170
  • explanation_gradients vertex-ai: add :UPLOAD integrated_gradients tests #12171
  • explanation_xrai vertex-ai: add :UPLOAD xrai_attributions test #12173
  • explanation_shapley vertex-ai: add :UPLOAD sampled_shapley_attributions test #12172

Release Note Template for Downstream PRs (will be copied)

See Write release notes for guidance.

[BBBmau]

latest commit fixes some errors from initial running the basic_upload test. After running locally it works as intended.

Next is breaking up the list of model upload types.

[BBBmau]

recent commit reflects updates from add-vertex-model-COPY

[BBBmau]

ready for review after cleanup

explanationSpec is commented out since the coming PRs are meant to comment out sections one by one.

[BBBmau]

some thoughts after getting upload_basic to pass, though there is some things to consider.

What's next is supporting the copy tests with upload in order to:

1. setup a model
2. copy the pre-made model in the same test in order to test the functionality of copy without needing a model to always be present since nightly tests update these and we need to be explicit with the model ID

[BBBmau]

Currently trying to resolve the diff we get on versionAliases where a user inputs their field [v2beta2] but get returned a diff on a second apply due to default_from_api returning [default, v2beta2]

although we could remove default_from_api this is necessary for copy to work since we don't input versionAlias yet we still expect a value from the model we want to copy.

In this case we can input the value but we should expect it to have default already as part of the config. I haven't been able to figure out a proper solution. A DiffSuppress was the first idea but because this is an array we are unable to use it (unless there's a hacky way that I don't know about. Leaving this for suggestions @SarahFrench

[BBBmau]

After some further investigation we may want to consider removing this for the time being, there is currently no way to include suppot of StateFunc which would help with supporting this (more can be found here). A Diff_suppress_func can't be applied due to lack of support on Lists.

I was almost convinced to leave it with ignore_read: true though this leaves us with a case where we don't receive the version_aliases when copying (e.g the target model contains v1beta1 but instead is set as null when attempting to copy)

Because of the reasons above it may be best to remove this field. Any thoughts on this would be appreciated.

[BBBmau]

this is included so that we can fill out the description field on create when wanting to upload.

As a reminder this diffsuppress was created because of copy not supporting description field on create. More info can be found in this thread. #12074 (comment)

[BBBmau]

recent commit supports UPLOAD for COPY tests, this allows us to not rely on a pre-existing model for source_model testing:

(base) ┌─(~/Dev/terraform-provider-google)────────────────────────────────────────────────────────────(mau@mau-JKDT676NCP:s083)─┐
└─(18:53:19 on vertex-ai-model-base ✹ ✭)──> envchain GCLOUD make testacc TEST=./google/services/vertexai TESTARGS='-run=TestAccVertexAIModel_'
TF_ACC=1 TF_SCHEMA_PANIC_ON_ERROR=1 go test ./google/services/vertexai -v -run=TestAccVertexAIModel_ -timeout 240m -ldflags="-X=github.com/hashicorp/terraform-provider-google/version.ProviderVersion=acc"
=== RUN   TestAccVertexAIModel_vertexAiModelUploadBasicExample
=== PAUSE TestAccVertexAIModel_vertexAiModelUploadBasicExample
=== RUN   TestAccVertexAIModel_postCreationUpdates
=== PAUSE TestAccVertexAIModel_postCreationUpdates
=== RUN   TestAccVertexAIModel_modelIdNotProvidedAtCreateTime
=== PAUSE TestAccVertexAIModel_modelIdNotProvidedAtCreateTime
=== RUN   TestAccVertexAIModel_modelIdProvidedAtCreateTime
=== PAUSE TestAccVertexAIModel_modelIdProvidedAtCreateTime
=== CONT  TestAccVertexAIModel_vertexAiModelUploadBasicExample
=== CONT  TestAccVertexAIModel_modelIdNotProvidedAtCreateTime
=== CONT  TestAccVertexAIModel_modelIdProvidedAtCreateTime
=== CONT  TestAccVertexAIModel_postCreationUpdates
--- PASS: TestAccVertexAIModel_vertexAiModelUploadBasicExample (23.28s)
--- PASS: TestAccVertexAIModel_modelIdNotProvidedAtCreateTime (34.93s)
--- PASS: TestAccVertexAIModel_modelIdProvidedAtCreateTime (35.33s)
--- PASS: TestAccVertexAIModel_postCreationUpdates (45.14s)
PASS
ok      github.com/hashicorp/terraform-provider-google/google/services/vertexai 46.573s

[BBBmau]

This update_encoder is necessary due to source_model being included in PATCH. The API Reference doesn't include any mention of source_model for patching existing models (such as for updating the description for a copied model)

Reference: https://cloud.google.com/vertex-ai/docs/reference/rest/v1/projects.locations.models/patch

[BBBmau]

This is ready for another review. Only thing to consider is whether we want to keep investigating solutions for versionAlias as discussed in #12075 (comment) @SarahFrench