Fix protobuf parser regex to handle scientific notation

[willpartcl]

Problem:
The regex pattern on line 241 could not parse floating-point values with scientific notation (e.g., 1.42109e-16), causing failures when loading 15 out of 17 IBM (ICCAD04) benchmarks.

Root Cause:
The pattern \-*\w+\.\*\/{0,1}\w*[\w+\/{0,1}\w*]* splits on word boundaries, breaking scientific notation into separate tokens:

• Input: "f: 1.42109e-16"
• Parsed as: ['f', '1.42109e', '-16']
• Caused float('1.42109e') to fail with ValueError

Solution:
Updated regex to explicitly match scientific notation first, then fallback to other patterns:

• Pattern: r'[-+]?\d+.?\d*[eE][-+]?\d+|[-]?\w+.?[\w/]*'
• Now parses: ['f', '1.42109e-16']
• Correctly handles positive/negative exponents

Testing:
Verified all 17 IBM benchmarks now parse successfully:

• ibm01-ibm18 (excluding ibm05) all load without errors
• Regex still correctly handles:
  • Regular floats (0.4, -0.4)
  • Integers (123)
  • Strings (TOP, BOTTOM)
  • Paths (foo/bar)
  • Scientific notation (1.42109e-16, 5.68434e+10)

Impact:
This fix enables plc_client_os to parse the full ICCAD04 benchmark suite without requiring Circuit Training's proprietary parser.