feat: include Bitrix24 task context in reports with secret redaction

[Meacue]

Problem this solves

Generated reports currently lack the task-level context that lives in Bitrix24 — the task description and the discussion in comments. Without it, a reviewer reading the report has to cross-check Bitrix24 manually to understand the why behind each branch and its commits, which defeats much of the point of an aggregated report.

At the same time, descriptions and comments in Bitrix24 routinely contain credentials pasted ad-hoc by developers (API keys, access tokens, DSN strings, private keys, and similar). Forwarding that text as-is to an LLM provider, embedding it into an exported Word document, or persisting it in our own storage would push those secrets outside the company perimeter. Any feature that brings Bitrix24 text into the report has to handle that risk up front — it cannot be a follow-up.

Proposed solution

• Pull task description and comments from the matched Bitrix24 task into the data used for report generation, so the narrative around each branch has the task's own context next to it.
• Before that text leaves the application (LLM call, generated Word document, persisted report), recognisable secrets in it are replaced with a typed placeholder, leaving the surrounding sentences intact and readable.
• The user can tell that redaction took place for a given report — they should not have to trust it silently.

Acceptance criteria

• Reports include relevant Bitrix24 task context (description + comments) alongside the matched branches.
• Common credential formats found in that context are masked before the text reaches the LLM, the generated .docx, or any persistent storage; the original secret is never written outside the redaction step.
• The user can see, per report, that redaction was applied (and how many items were masked).
• Redaction can be exercised end-to-end by a test that feeds known secret-shaped strings through the pipeline and asserts they never appear in the outputs.