fix(linkedin): preserve multiline safe-send messages

[gauravsaxena1997]

Summary

This fixes LinkedIn safe-send flattening multiline messages before they reach the LinkedIn composer.

The command previously validated --message through normalizeWhitespace(...) and then reused that normalized value for sending. That made messages with paragraphs, numbered answers, or bullet lists arrive at the composer as a single line. The composer then had no chance to preserve user-intended formatting.

Why this matters

linkedin safe-send is commonly used for careful human-approved outreach and replies. Those messages often need readable structure:

• short paragraphs
• numbered answers
• bullet lists
• screening-question responses

Flattening everything into one paragraph is hard to read and makes approved messages look less professional, especially for recruiter or lead replies.

What changed

• Preserve the raw --message string after validating that it is non-empty.
• Keep existing normalized comparisons for safety checks.
• Add a LinkedIn-specific multiline composer fill path for messages containing \n or \r.
• For multiline content, try paste-style contenteditable insertion first, then fall back to DOM multiline insertion with <br> nodes and input events.
• Keep the previous page.insertText(...) path for single-line messages.
• Keep all fail-closed recipient/thread/latest-message checks unchanged.

Safety model

This does not relax the existing write safety boundaries:

• dry-run remains the default
• exact thread URL is still required
• expected recipient is still verified
• optional latest-message text/hash guards still run before filling
• post-fill normalized composer text must still match the intended message
• send still requires explicit --send true

The main change is that validation no longer destroys formatting before insertion.

Validation

• npm run test:adapter -- clis/linkedin/safe-send.test.js
• npm run typecheck
• npm run test:adapter

Full adapter suite result: 379 files passed, 3717 tests passed.