Add parameter validation to prevent silent failures

[MaxGhenis]

Description

This PR adds parameter validation to the LanguageModel class to warn users about invalid or misspelled parameters that will be silently ignored.

Fixes #2184

Problem

Currently, EDSL accepts any parameter without validation, leading to silent failures when users:

• Make typos (e.g., temprature instead of temperature)
• Use unknown parameters (e.g., unicorn_mode=True)
• Use parameters not supported by their chosen service

Solution

Added a parameter validation system that:

1. Warns about unknown parameters
2. Suggests corrections for common typos
3. Validates parameters against service-specific lists
4. Uses warnings instead of errors for backward compatibility

Changes

• Added parameter_validator.py with validation logic and known parameters for each service
• Updated language_model.py to validate kwargs before applying them
• Added comprehensive tests in test_parameter_validation.py

Example

# This now generates a warning:
m = Model('gpt-4o-mini', temprature=0.7)
# UserWarning: Parameter validation issues:
#   'temprature' might be a typo. Did you mean 'temperature'?

# This also warns:
m = Model('gpt-4o-mini', unicorn_mode=True)
# UserWarning: Parameter validation issues:
#   Unknown parameters for openai: unicorn_mode
#   These parameters will be ignored and have no effect.

Testing

• Added comprehensive test suite covering:
  • Valid parameters (no warnings)
  • Unknown parameters (warnings)
  • Typo detection
  • Service-specific validation
  • Backward compatibility (parameters still set as attributes)

Backward Compatibility

• Uses warnings instead of errors
• Unknown parameters are still set as attributes (existing behavior preserved)
• Can be made stricter in future major version if desired