fix: deduplicate DataFrame index before asfreq in _apply_df_freq_horizon

Prevents ValueError: cannot reindex on an axis with duplicate labels
when naive-mpc-optim is called with an InfluxDB backend whose time
window (due to UTC offsets) spans two calendar days, producing ~2x
the expected data points. The resulting concat of PV + load forecast
Series can carry duplicate timestamps that cause asfreq() to fail.

Uses the same ~df.index.duplicated() pattern already present in
set_df_index_freq() and prepare_data() elsewhere in the codebase.

Author: hossamnagy

src/emhass/command_line.py

@@ -736,6 +736,7 @@ def _apply_df_freq_horizon(
         step = retrieve_hass_conf["optimization_time_step"]
         if not isinstance(step, pd._libs.tslibs.timedeltas.Timedelta):
             step = pd.to_timedelta(step, "minute")
+        df = df[~df.index.duplicated(keep="last")]
         df = df.asfreq(step)
     else:
         df = utils.set_df_index_freq(df)