PyDABs codegen: autogenerate core resource wiring - #6397

The PyDABs codegen writes databricks/bundles/<ns>/_models/*.py and each
namespace __init__.py, but those files carried no generated-file marker.

- Prepend a "# Code generated by pydabs-codegen. DO NOT EDIT." header in
  both writer paths (_write_code, _write_exports) and regenerate.
- Add a nested python/databricks/bundles/.gitattributes marking the
  generated _models trees and namespace __init__.py files as
  linguist-generated (core/ and resources/ stay hand-written).

Addresses review comments on #6335.

Co-authored-by: Isaac <no-reply@databricks.com>

Match all namespace __init__.py files and unset the hand-written core one,
instead of listing each generated namespace. New resource namespaces are
then marked automatically without touching .gitattributes.

Co-authored-by: Isaac <no-reply@databricks.com>

Wire the catalog resource into the core package: add_catalog, the catalogs
collection property, catalog_mutator, and the _ResourceType registration,
mirroring the other resources. The catalog model was already generated
(resources.Catalog is in RESOURCE_NAMESPACE) but never wired, so it was not
user-addable. Add a catalog case to test_resources.py.

Catalog is supported only on the direct deployment engine (there is no
terraform converter for it); this matches YAML-defined catalogs.

Co-authored-by: Isaac <no-reply@databricks.com>

Validate that a YAML-defined catalog and a Python-added catalog coexist,
serialize, and are mutated by catalog_mutator. Restricted to the direct
deployment engine, since catalogs have no terraform converter.

Co-authored-by: Isaac <no-reply@databricks.com>

Co-authored-by: Isaac <no-reply@databricks.com>

Replace the per-type Resources._jobs/_pipelines/... attributes with a
single self._resources dict keyed by plural name, populated by iterating
_ResourceType.all(). __init__ no longer needs a per-resource line; the
add_X methods and properties read/write self._resources[plural].
Behavior-preserving.

Co-authored-by: Isaac <no-reply@databricks.com>

Replace the per-type "for a job" / "for an alert" phrasing in the
duplicate-resource-name error with a uniform "for resource '<type>'".
This removes the grammatical-article special case so the wiring can be
generated uniformly. Drops the now-unused TestCase.article field.

Co-authored-by: Isaac <no-reply@databricks.com>

Add codegen/generated_wiring.py, which emits the per-resource wiring for
databricks.bundles.core into a new core/_generated/ package: the
_ResourceType registry (_resource_types.py), the Resources add_*/property
methods (_resources.py, a mixin), and the *_mutator decorators
(_resource_mutators.py). main.py calls it after the model-generation loop.

The core package __init__ is now generated too (static exports plus the
generated mutator exports); mark core/_generated/ generated in
.gitattributes and wipe it before regenerating.

The hand-written wiring in _resource_type.py / _resources.py /
_resource_mutator.py is removed in the next commit.

Co-authored-by: Isaac <no-reply@databricks.com>

Resources now inherits the generated _GeneratedResources mixin instead of
defining per-resource add_*/property methods; _ResourceType.all() returns
the generated tuple; and _resource_mutator.py keeps only the ResourceMutator
dataclass, with the decorators now generated. Tests import the mutators from
the public databricks.bundles.core instead of the internal module.

Behavior-preserving: the generated wiring reproduces what was hand-written.

Co-authored-by: Isaac <no-reply@databricks.com>

Address review feedback that the f-string-based generator was hard to
reason about. Each wired resource now gets its own _generated/<plural>.py
rendered from wiring_resource.py.tmpl (a string.Template with $-placeholders,
so the generated shape reads like real Python and there is no brace-escaping).
The generated _generated/__init__.py collects the per-resource mixins into
_GeneratedResources, exposes _all_resource_types(), and re-exports the
mutators. Per-resource files stay small as more resources are onboarded
instead of growing single aggregate modules.

Behavior unchanged: same generated API, tests green, regeneration idempotent.

Co-authored-by: Isaac <no-reply@databricks.com>

_bases() filtered out `object` but not underscore-prefixed private bases, so a
generated private base (e.g. _GeneratedResources introduced by the wiring
refactor) would surface in the golden as a false positive even though the public
contract is intact. Mirror _members()'s underscore filter. No-op for the current
golden on this branch, where Resources has no private base.

Co-authored-by: Isaac <no-reply@databricks.com>

…-guard' into sankalp-mittal/pydabs-autogen-wiring

_bases() filtered out `object` but not underscore-prefixed private bases, so a
generated private base (e.g. _GeneratedResources introduced by the wiring
refactor) would surface in the golden as a false positive even though the public
contract is intact. Mirror _members()'s underscore filter. No-op for the current
golden on this branch, where Resources has no private base.

Co-authored-by: Isaac <no-reply@databricks.com>

The core public-API snapshot doesn't exercise the CLI end to end, so it belongs
with the other pure-Python pydabs tests rather than in acceptance/. Move it to
databricks_tests/core/test_public_api.py, comparing against a committed golden
(regenerate with UPDATE_SNAPSHOTS=1). Gate on Python >= 3.11 (via skipif, and a
sys.version_info guard so pyright at 3.10 is happy) instead of pinning an exact
interpreter, since typing.get_overloads needs 3.11+ and the output is identical on
3.11/3.12/3.13. Removes the acceptance/bundle/python/public-api test.

Co-authored-by: Isaac <no-reply@databricks.com>

…-guard' into sankalp-mittal/pydabs-autogen-wiring

The per-resource files already came from wiring_resource.py.tmpl, but
_generated/__init__.py and core/__init__.py were still built from inline
f-strings in generated_wiring.py. Move those into generated_init.py.tmpl
and core_init.py.tmpl so all generated output is rendered from templates
(and highlights as Python). Generated output is unchanged. Addresses
reviewer feedback on #6397.

Co-authored-by: Isaac <no-reply@databricks.com>