[CI/helm] Enhancement of CI for helm charts

[yongfengdu]

Current CI for helm charts has 2 problems:
1). It use autodependency.enabled flag to handle the dependency, it works fine with one set of dependency, but with more flavors supported, this is hard to manage.
For Example:
data-prep depends on 2 services: Embedding and VectorDB.
For an AI app, it's common these services are shared with retriever-usvc, so it's the App's responsibility to pass the Embedding/VectorDB Endpoint to data-prep. However for CI to test the data-prep service itself, we'll need to bring up our own Embedding and VectorDB, autodependency.enabled provide the flexibility of deploying dependent services.
The original version only support TEI and Redis, so it works fine, but with more flavors support(Milvus as vectorDB, mosec as embedding endpoint), autodependency is not flexible enough to cover this, we need to fix it.
2). Current CI will run test on all different configurations(*values.yaml).
CI will run tests on all the values.yaml it found for the helm charts under test, and skip undesired one in the test script.
This might test more configurations than we expected. (cpu-values.yaml etc), and some values.yaml are added for reference only.
The better way is to explicitly specify CI covered cases, with well defined naming rules.
For example, we can use *-ci-values.yaml to let the CI know this values file is for CI to use.