The click-through navigation itself (company list, clickable dots in
the positioning matrix, routing via session_state) already existed.
This fills in the actual "deep-dive" content on that page, per the
issue's requirements: multi-year financials, not just the latest
period.

- metrics_engine.py: new build_historical_trend() builds a
  chronological (oldest-first) table of revenue, EBITDA, EBIT, net
  income and EBITDA margin per reported year for one company, reusing
  the already-cached raw data — no new network calls.
- app.py: render_company_page() now shows a Plotly line chart of that
  trend plus a formatted table, instead of the placeholder message.

Two issues found and fixed while testing live in a browser:
- Streamlit's st.dataframe didn't respect a pandas Styler's na_rep
  for missing values (showed the literal word "None"). Replaced with
  manually formatted strings built in a plain loop instead.
- The date-typed period column made Plotly draw a continuous date
  axis with odd monthly ticks between yearly data points. Added a
  plain year-label column and set the x-axis to category type.

Verified end-to-end with a headless-browser driver (click a company,
see the chart and table render with real cached data, no console
errors) before committing.

Full suite: 43 passed, 91% coverage on src/.

Closes #3