Align recommendation limitations with analysis maturity

[PooriaT]

Motivation

• Update deterministic recommendation wording so it no longer claims technical, fundamental, or news inputs are “not implemented” while still avoiding overstating feature maturity.
• Make limitations and per-item reasoning accurately reflect that technical indicators, market-condition analysis, fundamentals, and news summaries exist as optional/experimental, provider-dependent decision-support inputs while the deterministic rule set remains primarily snapshot/risk driven.

Description

• Replaced stale "not implemented" reasoning in _recommend_holding with conservative wording that states the deterministic status is derived from valuation, allocation, unrealized P/L, and risk warnings while optional technical/fundamental/news inputs exist but are not yet authoritative drivers. (see src/finwall/recommendations.py).
• Replaced the cash-deployment wording that claimed engines were unavailable with a clarification that cash deployment is based on valuation completeness, cash allocation, and risk warnings; optional technical/fundamental/news inputs are not used to generate ticker-level buy candidates in this rule set (see src/finwall/recommendations.py).
• Added a small helper _analysis_input_limitations() to centralize input-maturity wording and used it in _build_limitations() to replace the previous hard-coded "not implemented" limitation lines (see src/finwall/recommendations.py).
• Updated product-maturity documentation to reflect implemented optional/experimental analysis surfaces and to note that deterministic recommendations remain conservative and primarily snapshot/risk driven (README.md, docs/product-modes.md, docs/safety-limits.md).
• Added regression tests that assert stale phrases are absent and that new optional/experimental wording appears in report limitations and reasoning (tests/test_recommendations.py, tests/test_reports.py).
• No deterministic decision logic, thresholds, status assignment, cash-deployment math, report generation flow, API behavior, or risk-rule handling was changed.

Testing

• Ran lint/format checks: poetry run ruff check . and poetry run ruff format --check --line-length 88 ., both succeeded.
• Ran the full test suite: poetry run pytest, which passed (all tests including new regressions succeeded).
• New tests verify absence of stale phrases, presence of conservative optional-input wording, and preservation of existing data-quality and risk-limit warnings.

---

Codex Task