Feature request: incorporate cloudml-hypertune into the SDK and deprecate the standalone version

It's great that google-cloud-aiplatform has the ability to launch a hyperparameter tuning job.

In order to publish metrics for each trial, however, developers still have to use the **ancient** `cloudml-hypertune` library.

This leads to a poor developer experience:

- As a developer of hyperparameter tuning jobs on Vertex AI, I want to use only one SDK, the Vertex AI SDK
- seeing that the version of cloudml-hypertune on pypi is 0.1.0.dev6 is enough to make a developer wonder if the library is even supported anymore
- "cloudml" is an ancient name for Google Cloud's ML job running functionality, older than even "AI Platform"

Other oddities about the current setup:
- The 0.1.0 release from 2023 hasn't been published to pypi
- The 0.1.0dev6 version on pypi declares the most recent Python version as 3.5
- The cloudml-hypertune library `.travis.yml` file exposes what may be a pypi password, hopefully invalid by now:
https://github.com/GoogleCloudPlatform/cloudml-hypertune/blob/master/.travis.yml#L15

まず、スタンドアロンの cloudml-hypertune ライブラリと併せて、既存の Vertex AI SDK のハイパーパラメータ・チューニング機能を確認します。cloudml-hypertune の .travis.yml、特に参照されている認証情報の行を調査し、移行と非推奨化の範囲を明確にします。完了時には、サポートされている 1 つの SDK を通じて試行メトリクスを提供し、スタンドアロンパッケージの公開と Python バージョンに関する懸念に対処できるようにします。

索引モデルが issue の本文から書いたものです。