Fix wildcard regression in subscribe_current_values on v2 (closes #53)

[aki1770-del]

What

Restores branch-path support in subscribe_current_values on the v2 code path.
Since 0.5.1 preferred v2, calls like client.subscribe_current_values(['Vehicle'])
fail with NOT_FOUND: the v2 Subscribe RPC only accepts fully-qualified leaf
paths. v1 Subscribe accepted branch paths natively, so this is a regression for
existing users.

Fixes #53.

How

• On the v2 path, catch NOT_FOUND from the first Subscribe attempt.
• Expand the input paths via ListMetadata(root=path) — the v2-native way to
  enumerate signals under a branch — and retry Subscribe with the resulting
  leaves.
• Leaf paths stay on the fast path: they never trigger the retry, so no extra
  RPC in the common case.
• Trailing .* (e.g. Vehicle.Cabin.*) is stripped before lookup so both the
  v1 idiom ['Vehicle'] and the explicit wildcard form work.
• The existing UNIMPLEMENTED → v1 fallback is preserved for old databrokers.

Scope

Strictly subscribe_current_values (sync + async). subscribe_target_values
has the same shape but is out of scope here — happy to file a symmetric fix in
a follow-up if the approach looks right.

Existing behavior audit

• v2_subscribe implementation: kuksa-client/kuksa_client/grpc/aio.py:537 —
  passes paths directly into SubscribeRequest; no client-side expansion.
• _prepare_v2_list_metadata_request: kuksa-client/kuksa_client/grpc/__init__.py:907
  already builds the ListMetadata request shape. Reused here.
• val.proto semantics for ListMetadata: "Shall correspond to a VSS branch,
  e.g. Vehicle, Vehicle.Cabin. Metadata for all signals under that branch
  will be returned." (kuksa/val/v2/val.proto)
• v1 get_metadata still works with wildcards (per the issue reporter) — which
  is why this is a clear regression from the v2 switch, not a missing feature.

Prior art / related

• Wildcard usage for subscribe_current_values not working anymore since 0.5.1 #53 — original bug report (psch0rr)
• Migration to v2 timeline/plan? #41 — Schildt's v2 migration timeline issue (establishes v2 as strategic direction)
• Extend subscribe_target_values to use v2 API #50 — Extend subscribe_target_values to use v2 API (BjoernAtBosch, merged)
• Fix set_current_values using kuksa.val.v2 API #49 — Fix set_current_values using kuksa.val.v2 API (BjoernAtBosch, merged)

Tests

• test_subscribe_current_values — unchanged, still passing (leaf-path fast path)
• test_subscribe_current_values_branch_path_expansion — new; asserts that on
  NOT_FOUND the client retries v2_subscribe with expanded leaves
• Full suite: 274/274 pass locally

Not tested (would appreciate reviewer input)

• Integration against a live databroker v0.6.0 — done by reporter per issue;
  happy to add an integration test if the project has a pattern for that.