feat(databricks): add explicit discovery workflow

[ElliotSun]

Summary

Add the missing Databricks read-side discovery workflow for M1 while preserving the existing platform observation boundary.

This PR is scoped to Discovery + Observation only. Contract import is an already implemented SemaPact capability and is not implemented, refactored, or exercised here.

Read-side boundaries

Initialized WorkspaceClient
        │
        ├── Discovery
        │     tables.list(...)
        │        ↓
        │     table FQNs
        │
        └── Observation
              tables.get(...)
                 ↓
          ObservedPlatformState

The existing contract-import path remains separate and unchanged. Discovery and observation must not depend on or invoke it.

Changes

• add semapact.platforms.databricks.discover_databricks_tables(...) backed by WorkspaceClient.tables.list(...);
• return deterministic, deduplicated fully-qualified table names only;
• require an initialized client for discovery and keep auth resolution outside the workflow;
• keep discovery free of observation, ODCS import, lifecycle, and governance dependencies;
• preserve observe_databricks_table(...) behavior unchanged;
• rely on dependency-isolation tests to keep discovery/observation independent from importer and governance layers;
• extend the fresh semapact[databricks] smoke to verify the official SDK read-side surfaces only.

Scope semantics

Discovery
→ Which Databricks assets are available in this scope?
→ Output: table FQNs

Observation
→ What state does Databricks currently report for a selected asset?
→ Output: ObservedPlatformState

Discovery does not imply observation. Observation does not imply contract import. Neither operation mutates governed ODCS state.

Out of scope

• contract import implementation/refactoring/testing — already implemented;
• new discovery/observation CLI commands;
• authentication mechanisms;
• reconciliation/drift classification;
• observation fingerprints;
• rich evidence/lineage;
• release/deployment behavior.