chore(tests): normalize recording IDs and timestamps to reduce git diff noise (#3676)

IDs are now deterministic hashes based on request content, and
timestamps are normalized to constants, eliminating spurious changes
when re-recording tests.

## Changes
- Updated `inference_recorder.py` to normalize IDs and timestamps during
recording
- Added `scripts/normalize_recordings.py` utility to re-normalize
existing recordings
- Created documentation in `tests/integration/recordings/README.md`
- Normalized 350 existing recording files

tests/integration/recordings/responses/6412295819a1.json

@@ -16,7 +16,7 @@
     "body": {
       "__type__": "openai.types.completion.Completion",
       "__data__": {
-        "id": "cmpl-865",
+        "id": "rec-6412295819a1",
         "choices": [
           {
             "finish_reason": "stop",
@@ -25,7 +25,7 @@
             "text": "Blue.\n\nMy answer is \"blue\" because it's a classic completion of the traditional nursery rhyme poem:\n\n\"Roses are red, violets are blue\"\n\nThis sentiment suggests that an unseen suitor from the first half of the line has given or will give the speaker roses."
           }
         ],
-        "created": 1759441353,
+        "created": 0,
         "model": "llama3.2:3b-instruct-fp16",
         "object": "text_completion",
         "system_fingerprint": "fp_ollama",