bufio.Scanner 10MB token limit may truncate large transcript lines

[khaong]

Problem

parseTranscriptFromLine in agent/claudecode/transcript.go uses bufio.Scanner with a 10MB max token size. Claude Code tool_use JSONL lines can embed large file contents, and if a single line exceeds that limit, parsing will error.

This affects all callers:

• ExtractAllModifiedFiles
• CalculateTotalTokenUsage
• Other transcript parsing functions

Impact

If a transcript line exceeds 10MB, file extraction and token usage calculation will fail. In ExtractAllModifiedFiles, the error triggers a fallback to main-transcript-only extraction, potentially missing subagent files.

Suggested Fix

Consider switching to a bufio.Reader-based parser (like ClaudeCodeAgent.ExtractModifiedFilesFromOffset) that doesn't have fixed token limits.

Context

Identified during review of #323 (ENT-297 fix).

[gtrrz-victor]

Closed by: #346