feat: updating Tensorboard related code to use v1

[yfang1]

Fallout of #849

[morgandu]

Left a comment which applies to all the changes. So the rule of thumb is:

if your expected version aligns with the DEFAULT_VERSION in the compact module, no need to specify version i.e. from google.cloud.aiplatform.compat.types import tensorboard_data; but if your expected version is different than the DEFAULT_VERSION, import the particular version. i.e. from google.cloud.aiplatform.compat.types import tensorboard_data_v1beta1.

[morgandu]

Sorry I probably missed the test_uploader.py in the previous review. For unit tests, we don't rely on the compat module, instead we specify the expected version. Because we don't have tests when compat may or may not select different version within different resource class methods, this way we can ensure class methods calling compat returning the right version we expect.

We only discussed this when compat was introduced, but did not enforce the implementation across all other module, so there is some inconsistency in the code base. But prefer start to enforce it for the reason mentioned above.