Rename get_default_config to get_default_metrics in redis_cloud/redis_enterprise checks

[philjlee]

What

Renames the get_default_config() method on RedisCloudCheck and
RedisEnterpriseCheck to get_default_metrics().

Why

Both checks defined their own get_default_config() returning a list of
metric groups. A newer datadog-checks-base adds its own
get_default_config() hook to OpenMetricsBaseCheckV2, used internally by
get_config_with_defaults() to seed scraper config from the check's
generated config_models defaults (expects key/value pairs). Because
pyproject.toml pins datadog-checks-base>=37.20.0 with no upper bound,
CI resolves the latest version, whose base class method gets shadowed by
each check's same-named override — get_config_with_defaults() then calls
the check's list-returning method instead and dict() raises:

ValueError: dictionary update sequence element #0 has length 65; 2 is required

Confirmed pre-existing and unrelated to any other open PR: reproduced on an
unmodified checkout of master via uvx --python 3.13 ddev@18.0.0 -x test redis_cloud — the failure is byte-for-byte identical. It's only visible in
CI when a PR happens to touch files inside redis_cloud/ or
redis_enterprise/, since TARGET=changed scopes the test matrix to
touched directories.

How

Renamed the check-specific method (and its one call site in
_parse_config) in both integrations from get_default_config to
get_default_metrics, which no longer collides with the base class's
config-defaults hook.

Verified locally with ddev@18.0.0 (the version CI's test workflow
resolves): redis_cloud 5/5 passing, redis_enterprise 7/7 passing, lint
clean.

[philjlee]

Superseded — combined into #3127. Both PRs touched the same redis_cloud/redis_enterprise directories, and under CI's TARGET=changed scoping neither could pass independently of the other (whichever was checked standalone would fail on the other's pre-existing bug). Folded the get_default_config rename into #3127 alongside the drift fix.