Preserve leading-zero JSON keys in nested items

[harsh-thakkar7]

Summary

A nested JSON item with a leading-zero literal, e.g. 01=x, was tokenized as a NUMBER (Python's int('01') → 1) and emitted as the key "1" — silently rewriting the key and colliding with a real 1=x item, losing data:

$ http example.com/post '01=x' '1=y'
{"1": "y"}     # '01' became '1' and overwrote '1'!

01 is a valid JSON key; JSON numbers cannot have leading zeros, so coercing it to an integer is lossy.

Fix

In httpie/cli/nested_json/parse.py, only treat a literal as a number when it is canonical — i.e. its str() form round-trips exactly (int(value) without leading zeros / missing sign). Non-canonical literals remain TEXT keys:

def canonical_int(value: str) -> int:
    if not value:
        raise ValueError('Not an int')
    converted = int(value)
    if str(converted) != value:
        raise ValueError('Not a canonical int')
    return converted

$ http example.com/post '01=x' '1=y'
{"01": "x", "1": "y"}    # keys preserved, no collision

Canonical integers are unaffected (kids[0], kids[10], ... still produce numeric array indices).

Tests

• ['01=x', '1=y'] → {'01': 'x', '1': 'y'} — fails before this change (key collapsed to 1).
• ['kids[0]=Thelma', 'kids[10]=Ashley'] → indices still numeric.

pytest -q tests/test_json.py tests/test_cli_utils.py   # 298 passed

[harsh-thakkar7]

Closing this — I don't think this change is mature enough to land right now. Thanks for the project!