fix: reset self.df at start of compute_metadata to prevent DuplicateError

compute_metadata mutates self.df by adding timezone and local_datetime
columns. Calling it a second time on the same instance caused a
DuplicateError in the Polars join (_compute_timezones). Fix: re-read
the CSV at the start of each call alongside the existing cache resets.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: koito19960406

src/zensvi/metadata/mly_metadata.py

@@ -821,6 +821,9 @@ def compute_metadata(
         # Clear per-call caches so repeated compute_metadata calls don't use stale data
         self._joined_daynight_cache = None
         self._seasons_grouped_cache = None
+        # Reset self.df from disk so timezone/other columns added in a previous call
+        # don't cause DuplicateError when compute_metadata is called more than once.
+        self.df = pl.read_csv(self.path_input)
 
         # check indicator_list and pre-compute metadata, e.g., timezone and local datetime
         if (