fix(benchmark): compare agents against DJIA total return (DIA baseline)

[shuxue6662-a11y]

Fixes #1434

Summary

The tutorial compares agents that trade dividend-adjusted (total-return) prices against a ^DJI baseline, which is a price-return index. The two sides therefore use different return definitions, and the reported gap includes the DJIA dividend stream (~+1.92%/yr over 2018-2025, per the issue's measurement) regardless of agent behavior.

This PR makes the baseline total-return too:

• examples/FinRL_StockTrading_2026_3_Backtest.py: download DIA (the DJIA-tracking ETF) with auto_adjust=True instead of ^DJI, with a comment explaining the convention.
• finrl/plot.py: change the backtest_plot default baseline_ticker from ^DJI to DIA. get_baseline already routes through YahooDownloader, which adjusts close prices for distributions, so the plotted baseline now matches the agent's total-return account value.
• README.md: clarify that the DJIA baseline is total return via the DIA ETF.

Notes

• DIA carries a small expense ratio (~0.16%/yr), so it slightly understates a pure DJIA total-return index; it is used because it needs no new data source (as discussed in Benchmark comparison: agent trades dividend-adjusted prices while the DJIA baseline is price-return only #1434).
• The agent input convention (Adj Close) is intentionally unchanged; only the comparison side is fixed.

Validation

• black --check on both modified Python files: pass
• flake8 on both modified Python files: no new issues (remaining warnings are pre-existing)
• Live yfinance re-download was rate-limited on this machine; the +1.92%/yr convention estimate in Benchmark comparison: agent trades dividend-adjusted prices while the DJIA baseline is price-return only #1434 was measured independently over 2018-2025.

[shuxue6662-a11y]

Hi @spencerromo, this PR fixes the benchmark convention mismatch described in #1434 (agents trade total-return prices while the ^DJI baseline is price-return only). The baseline now uses DIA with adjusted prices in both the tutorial and plot.py. Happy to adjust if you prefer documenting the convention instead.