source-google-play: complete initial development

[Alex-Bair]

Description:

This PR's scope includes:

• Adding a convenience BaseCSVRow class in the CDK to make transforming null value representations in CSVs easier and more consistent across connectors.
• Completing development for source-google-play that was started in source-google-play: new connector #3091.
  • Most of the changes were ones I anticipated on the initial implementation. See the commit message for more details, but the high level summary is:
    • reviews can be incrementally captured within each file since each row has an "updated_at" type field.
    • Column names are inconsistently cased, which would inevitably cause issues downstream. The predominant casing used for fields is Title Case, so I'm transforming all fields to abide by that convention.
    • We don't need to add "Row Number" to statistics since there already exists a sufficient composite primary key within each row.

Workflow steps:

(How does one use this feature, and how has it changed)

Documentation links affected:

Documentation should be created for source-google-play.

Notes for reviewers:

Tested on a local stack. Confirmed:

• All expected documents are captured.
• Backfills complete as expected.
• Incremental tasks capture as expected, albeit I only watched it for an hour or so.
  • These CSVs are likely updated once per day, so I'll keep an eye on production captures to make sure their incremental replication works as expected over multiple days.

---

This change is 

[JustinASmith]

LGTM % one question

[JustinASmith]

Does the model need the ConfigDict set up with serialize_by_alias=True so that the serialized JSON uses the alias Package Name consistently. https://docs.pydantic.dev/latest/api/config/#pydantic.config.ConfigDict.serialize_by_alias says that version 2 of Pydantic this is False by default, so model.model_dump() would presumably give #> {'package_name': 'some.package.com'}.

As a dummy example I tried the following:

from pydantic import BaseModel, Field, model_validator, ValidationInfo
from typing import Any, ClassVar

PACKAGE_NAME_FIELD = "Package Name"

class Statistics(BaseModel, extra="allow"):
    suffix: ClassVar[str | None] = "overview"
    primary_keys: ClassVar[list[str]] = ["/Date", f"/{PACKAGE_NAME_FIELD}"]
    date: str = Field(alias="Date")

    @model_validator(mode="before")
    @classmethod
    def _normalize_field_names(cls, data: dict[str, Any], info: ValidationInfo) -> dict[str, Any]:
        normalized_data: dict[str, Any] = {}
        for key, value in data.items():
            normalized_key = key.title()
            normalized_data[normalized_key] = value

        return normalized_data
        
s1 = Statistics(
    Date="2023-10-01",
    **{PACKAGE_NAME_FIELD: "com.example.app"}
)

s1.model_dump()

#> {'date': '2023-10-01', 'Package Name': 'com.example.app'}

[JustinASmith]

Scratch that! I just realized the CDK handles this by default: https://github.com/search?q=repo%3Aestuary%2Fconnectors%20path%3A%2F%5Eestuary-cdk%5C%2F%2F%20by_alias&type=code

[JustinASmith]

@Alex-Bair Sorry I just thought of one more question. Was there a reason these acmeCo/*.yaml were removed?

[Alex-Bair]

If we don't have a capture snapshot test, the various acmeCo/*.yaml files aren't needed since they're only used when we run the flowctl preview command. When we don't have a capture snapshot test, we can just leave bindings: [] (empty) in test.flow.yaml, and the spec & discover snapshot tests still work fine.

I haven't been extremely diligent to avoid adding these files when they're not needed, but I'm trying to keep it in mind more moving forward to avoid some clutter.