feat: migrate remaining consumers to new Plugin Architecture

- Add compatibility functions to tts_plugin/compat.py:
  - get_metadata(): returns TTSBackendMetadata with voices
  - is_registered_backend(): checks if plugin is loaded
  - resolve_backend_for_voice(): resolves backend for voice spec
  - get_default_voice(): gets default voice for backend

- Update tts_plugin/__init__.py to export new functions

- Migrate all consumers from old tts_backend_registry:
  - WebUI: conversion_runner, debug_tts_runner, routes/api, routes/utils/*
  - PyQt UI: gui, predownload_gui, voice_formula_gui
  - Voice utilities: voice_cache, voice_formulas, voice_profiles
  - Other: subtitle_utils, utils, predownload_gui (root)

- Update tests to use new plugin architecture

Old architecture remains intact as fallback.

abogen/webui/routes/utils/form.py

@@ -7,7 +7,7 @@ from flask.typing import ResponseReturnValue
 
 from abogen.webui.service import PendingJob, JobStatus
 from abogen.webui.routes.utils.service import get_service
-from abogen.tts_backend_registry import is_registered_backend
+from abogen.tts_plugin.compat import is_registered_backend
 from abogen.webui.routes.utils.settings import (
     load_settings,
     coerce_bool,
@@ -33,7 +33,7 @@ from abogen.webui.routes.utils.common import split_profile_spec
 from abogen.utils import calculate_text_length
 from abogen.voice_profiles import serialize_profiles, normalize_profile_entry
 from abogen.chunking import ChunkLevel, build_chunks_for_chapters
-from abogen.tts_backend_registry import get_default_voice
+from abogen.tts_plugin.compat import get_default_voice
 from abogen.speaker_configs import get_config
 from abogen.kokoro_text_normalization import normalize_roman_numeral_titles
 from dataclasses import dataclass