chore: skip naming analysis during initial type discovery

[ohmayr]

The generator builds its API model in two stages:

1. Discovery: Identifying all messages, enums, and services.
2. Generation: Resolving names and rendering Python code.

This change bypasses the expensive naming collision analysis during the Discovery stage. Since naming is fully re-calculated during the Generation stage, performing it twice is redundant and significantly slows down the build for large APIs.

Verification:

• zero-diff in output (verified via goldens).