Support _request_timeout in dynamic client watch()

[amel-ch]

Problem

When using the dynamic client’s watch() method:

api = self._client.resources.get(
    api_version=self._api_version,
    kind=self._kind
)

for event in api.watch(
    timeout=server_timeout_seconds,
)

there is currently no way to configure a client-side request timeout (_request_timeout).

Current Workaround

To enforce a client-side timeout, we have to bypass Resource.watch() and directly use Watch().stream():

for raw_event in Watch().stream(
    api.get,
    serialize=False,
    deserialize=False,
    timeout_seconds=self._server_timeout_seconds,
    _request_timeout=self._client_timeout_seconds,
)

Issue

The dynamic client’s watch() method internally relies on Watch().stream() but does not expose _request_timeout as a parameter: https://github.com/kubernetes-client/python/blob/master/kubernetes/base/dynamic/client.py#L203

This creates:
Inconsistent API ergonomics
Unnecessary duplication of logic for users needing client-side timeouts

Proposed Solution

Extend Resource.watch() to accept an optional _request_timeout parameter and pass it through to the underlying Watch().stream() call.

Example:

api.watch(
    timeout=server_timeout_seconds,
    _request_timeout=(connect_timeout, read_timeout),
)