tech-debt: standardize JSON library usage (json vs ujson) across the SDK

[dgarros]

Need

The SDK uses two JSON libraries interchangeably with no documented rule for
which to use when. Across infrahub_sdk/, ~16 modules use ujson (the apparent
house standard) and ~7 use stdlib json; playback.py and utils.py import
both. This makes the codebase harder to reason about and creates a latent
exception-handling hazard: ujson raises its own ujson.JSONDecodeError and
stdlib raises json.JSONDecodeError, and the two do not catch each other. Today
every decode/except pair happens to be internally consistent, but nothing
enforces that — a future ujson.loads guarded by except json.JSONDecodeError
(or the reverse) would silently fail to catch a malformed-input error.

Context

Concrete smells found while surveying usage:

• playback.py serializes with json.dumps (:52) but reads the recorded
  files back with ujson.load (:57). This works today (both speak valid JSON)
  but the asymmetry is a smell and imports both libraries in one module.
• utils.py also imports both json and ujson.
• The decode/except pairs are currently consistent everywhere
  (e.g. pytest_plugin items pair ujson.loads with except ujson.JSONDecodeError;
  infrahub_filters.py and parsers.py pair stdlib json.loads with
  except json.JSONDecodeError), so this is a maintainability/consistency
  concern rather than an active bug.

Note: an earlier framing suspected utils.py:97 (decode_json) of catching the
wrong exception for a ujson failure — that turned out to be a false alarm. That
line decodes via httpx's response.json() (stdlib-backed), so its
except json.decoder.JSONDecodeError is correct.

Affects: Python SDK.