fix(pixel-inference): reject boolean value for integer seed parameter

[vaibhavsrv]

Why this matters

In ods/extensions/services/pixel-inference/app/main.py, _prepare(payload, grant) lists seed in the allowed FIELDS mapping but does not validate its data type. In Python, bool is a direct subclass of int, so type checks such as isinstance(val, int) evaluate to True for True and False. When a client sends {"seed": true} or a non-integer float/string (e.g. "42"), the unvalidated value passes to the model engine, where boolean flags or floats alter backend pseudo-random number generator seeding unexpectedly or trigger runtime type errors.

This change introduces strict integer validation for seed in _prepare. If present and non-null, it requires type(seed) is int and type(seed) is not bool, raising ShareError(400, 'invalid_seed') for booleans, strings, or floats. Standard integer seeds and omitted/null seeds continue to function cleanly.

Validation

• Tested baseline reproduction against unpatched code: passing {"seed": True} or {"seed": 42.0} bypassed _prepare without error.
• Tested post-fix behavior: boolean and non-integer seed values are rejected with HTTP 400 and code invalid_seed; valid integer seeds pass without modification.
• Telemetry statement: "Pixel-inference suites: 3 passed. New-test Ruff, py_compile, and diff checks pass; new regression wired into Linux CI."

Overlap check

• fix(pixel-inference): reject negative or out-of-range temperature in prepare #5940 (open): Validated temperature bounds in pixel-inference.
• fix(pixel-inference): validate presence_penalty and frequency_penalty float bounds #5941 (open): Validated penalty bounds in pixel-inference.
• fix(pixel-inference): validate stop parameter type as string, array of strings, or null #5942 (open): Validated stop types in pixel-inference.
• fix(pixel-settings): reject active token limits exceeding provider capacity #5696 (merged): Clamped active token limits in pixel_settings/contract.py.
  This PR solely hardens seed type validation in pixel-inference/app/main.py and does not overlap with any concurrent PRs.

Risk / AI disclosure

AI-assisted defect discovery, implementation, and regression verification. This strengthens request boundary checking for inference random seeds and does not modify running infrastructure, models, or configurations. Independent human review remains a gate.

Follow-up integration evidence

Composed cleanly on top of #5569, #5871, and #5875 at HEAD without conflicts. Production and test diffs passed together; adjacent pixel-agent and pixel-settings checks remain intact.
Backlog composition was local-only (production/test diffs, excluding workflow/Makefile wiring); it is not an upstream merge or independent human approval. Declared live-review gates remain open.