Add fickling.loads() public API function

[dhalf]

Add fickling.loads() Public API Function

Summary

This PR adds a new public API function fickling.loads() that provides safe loading of pickle data from bytes, completing the API to match Python's pickle module.

Problem

The fickling library currently only exposes fickling.load(file) for loading from file objects. While hooks for pickle.loads() exist internally, there's no corresponding public fickling.loads(data) API function for loading
from bytes data. This creates an incomplete API compared to Python's pickle module, which provides both load() and loads().

Solution

Added fickling.loads() function that:

• Accepts bytes data (matching pickle.loads() signature)
• Wraps bytes in io.BytesIO and delegates to existing load() function
• Provides the same security analysis and safety guarantees as fickling.load()
• Supports all the same parameters (max_acceptable_severity, print_results, json_output_path, etc.)

Changes

Core Implementation

• fickling/loader.py: Added loads() function (42 lines) after existing load() function
• fickling/init.py: Exported loads alongside load and updated documentation comment

Tests

• test/test_loads_api.py: New test file with 4 comprehensive test cases:
  1. test_loads_safe_data - Verifies safe pickle data loads correctly
  2. test_loads_unsafe_data - Verifies unsafe data is rejected with UnsafeFileError
  3. test_loads_matches_pickle_signature - Tests compatibility with pickle.loads() arguments
  4. test_loads_with_custom_severity - Tests custom severity level support

Design Rationale

• Code Reuse: Delegates to existing load() function to avoid duplication and maintain consistency
• Simple Implementation: Just wraps bytes in io.BytesIO - all security logic remains in load()
• API Completeness: Matches Python's pickle module interface for easy migration
• Backward Compatible: No changes to existing functionality

Testing

All new tests pass:

$ pytest test/test_loads_api.py -v
============================== 4 passed in 0.02s ===============================

No regressions introduced - all previously passing tests still pass.

Usage Example

import pickle
import fickling

# Create pickle data
data = pickle.dumps([1, 2, 3])

# Before: Had to manually wrap in BytesIO
import io
obj = fickling.load(io.BytesIO(data))

# After: Direct drop-in replacement for pickle.loads()
obj = fickling.loads(data)

Benefits

1. Complete API - Now provides both load() and loads() like pickle
2. Better Ergonomics - No need to manually wrap bytes in BytesIO
3. Easy Migration - Drop-in replacement for pickle.loads()
4. Consistent Safety - Same security guarantees as fickling.load()

[dhalf]

#84

[thomas-chauchefoin-tob]

I fixed the logic of the last test, let's merge this one as well.