docs: bring specs back in line with the code

[mohammadp1001]

Section B of the round-3 audit: documentation-only drift. No behaviour changes, no code touched.

data-model.md

• Config model still had ticker1/ticker2; it is tickers: list[str] with any number supported. extended_hours was missing entirely. session_money_budget now documents the > 0 rule (fix: session_money_budget has no positive-value check (config-schema.md) #170).
• candle_timeframe was missing 10min and 15min.
• Session id was documented as session-{random_seq}; it is a zero-padded ordinal, run-1/session-0001 - which is also what spec.md Clarification Q4 says, so the two specs contradicted each other.
• execution_result values were EXECUTED/SKIPPED_BUDGET/SKIPPED_MARKET_CLOSED/SKIPPED_API_ERROR; the code writes EXECUTED/HOLD/FAILED, with the specific reason in ORDER_FAILED telemetry.
• trailing_stop_high_watermark was documented as NULL for non-Momentum; it is seeded to entry_price for every position (trailing_stop_high_watermark is never initialised to entry_price #130).
• "zero, one, or two Positions per Session" predated multi-ticker support.

contracts/agents.md, research.md

• The execution-telemetry known gap was closed by feat: emit execution telemetry on every order path (#132) #141; both files still described it as open. Replaced with what is actually emitted, including remaining_session_budget.
• research.md said the execution agent subclasses google.adk.BaseAgent and overrides _run_async_impl. It is a plain Python class with model = None; recorded why the ADK subclass was dropped.
• The "EVALUATION_FAILED missing from EVENT_TYPES" note was also stale.

contracts/report-context.md

• Named _run_session; the method is _process_session.
• Omitted ticker_details[].execution_result, which is populated - the note that "execution_result is always None" was only true of the top-level field.
• Said the feedback agent parses id="investment-thesis"; it is ticker-scoped as investment-thesis-{ticker} (Feedback agent extracts the wrong ticker's thesis from multi-ticker reports #134), with a documented fallback for older reports.

contracts/cli.md

• Documented 3 of 6 commands. version, verify-telemetry and reset were missing.
• --timeframe listed 3 of 5 accepted values.
• --budget said "0 or negative = no limit"; negative is now rejected (fix: session_money_budget has no positive-value check (config-schema.md) #170).
• Sample output used the old random-suffix session ids.

spec.md

• Said the investigation agent is invoked "per unblocked ticker"; it is invoked once with all unblocked tickers.

tasks.md

Left as a historical task record rather than rewritten, with a note at the top pointing at the two places it has been overtaken: the ordinal evaluation window (replaced by an absolute UTC deadline in #122) and the unscoped thesis id (#134).

609 passed in 139.57s (0:02:19)
Required test coverage of 100% reached. Total coverage: 100.00%
ruff check alphoryn/ tests/ -> All checks passed!