Fix InfluxDB hook methods failing when called before get_conn

[jayamanikharyono]

InfluxDBHook.write() and the five bucket/organization methods read self.client directly, but that attribute stays None until something calls get_conn(). Any Dag that calls them on a fresh hook fails with AttributeError: 'NoneType' object has no attribute 'write_api'.
They now take the client from self.get_conn(), which is what query() and query_to_df() in the same class have always done. get_conn() already returns a cached client, so this opens no additional connections.

Three further defects in the same hook are fixed alongside it:

• get_uri() fell back to port 7687 — Neo4j's Bolt port — when the connection carried no explicit port, so a connection created without one pointed at an unreachable address. It now defaults to 8086 for http and 443 for https, matching InfluxDB3Hook.get_uri() in the same provider.
• create_bucket() accepts a retention_rules argument and then passed a hardcoded None to the client, silently discarding whatever the caller supplied.
• The system Dag reads influxdb_hook.org_name, which has not existed since the constructor moved to self.extras. It also passed an organization name into create_bucket()'s org_id parameter, so it now resolves the name to an
  ID first.

---

Was generative AI tooling used to co-author this PR?

• Yes (please specify the tool below)

[fat-catTW]

LGTM!
Thanks for contribution.

[eladkal]

cc @arpitrathore can you take a look?

[arpitrathore]

LGTM!
Thanks for fixing the uninitialized client across the hook methods, the default port (8086/443, matching InfluxDB3Hook), and create_bucket silently dropping retention_rules. Good call removing the hand-assigned self.client from the tests, that's what turns them into actual regression tests.