fix: validate full amount string before parsing

[AtakanGs]

Summary

Fixes #38 by validating the entire normalized amount string before converting it to a number.

parseFloat() accepts a valid numeric prefix and ignores trailing invalid characters, so inputs such as 10abc or 1.25xyz were previously accepted as 10 and 1.25.

This change:

• validates the complete normalized string with a decimal-number pattern
• switches conversion from parseFloat() to Number() after validation
• keeps the existing positive/finite amount check
• preserves supported formatting normalization for currency symbols, commas, whitespace, parentheses, and Unicode minus

Behavior

Valid examples continue to work:

• $1,234.56 → 1234.56
• 1,234.56 → 1234.56
• 1.25 → 1.25
• .5 → 0.5

Malformed values are now rejected:

• 10abc
• 1.25xyz
• 5USDCx

Testing

The repository does not currently include a test script/framework for this utility, so I verified the parsing behavior directly with Node and also ran git diff --check successfully.

Closes #38.