Add confidence scores to `ProofreadCorrection` output

[yashrajbharti]

It would be helpful for each ProofreadCorrection to include an optional confidence field (e.g., a float between 0 and 1) indicating the model's certainty in its suggestion. This can support better UX decisions, like when to auto-apply a correction or show it as optional or filter out.

Example 1:

"She going to the store."

{
  "startIndex": 4,
  "endIndex": 10,
  "correction": "is going",
  "type": "missing-words",
  "explanation": "The verb 'is' is missing to form the correct present continuous tense.",
  "confidence": 0.93
}

Example 2:

"He walked the building."

{
  "startIndex": 10,
  "endIndex": 10,
  "correction": "towards",
  "type": "preposition",
  "explanation": "A preposition is needed to indicate the relationship between 'walked' and 'the building'.",
  "confidence": 0.55
}

(towards, into, past, around are all valid)