fix: avoid retrying callable backends on TypeError

[OBaruch]

Description

Prevents CallableClient from retrying a backend when the backend itself
raises TypeError.

The previous implementation used a broad except TypeError to fall back from
the full keyword call to the minimal (messages) signature. That also caught
errors raised inside the backend, causing a second invocation, duplicate side
effects, and loss of the original failure context.

The client now inspects and binds the callable signature before invocation. It
keeps support for both documented callable forms while ensuring the backend is
executed at most once per request. Unsupported signatures are rejected before
execution.

Type of change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that changes existing behavior)
• Documentation only

Checklist

• CLA signed when prompted by CLA Assistant
• Commit follows Conventional Commits
• Tests added for the fixed behavior
• Changelog updated
• No secrets, credentials, internal URLs, or proprietary content included

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[OBaruch]

I have read the CLA Document and I hereby sign the CLA