fix: strip '*' prefix from sha256sum filenames in checksums parser

The checksums.txt files on GitHub Releases use the standard
sha256sum format where binary-mode files have a '*' prefix
(e.g. '70c4... *firefox-150.0.1-stealth-linux-x86_64.tar.gz').

_parse_checksums() was using parts[-1] as the filename key
directly, which included the '*', causing the lookup for a
clean-filename asset (e.g. 'firefox-150.0.1-stealth-linux-x86_64.tar.gz')
to fail with 'no SHA256 in checksums.txt'.

Fix strips leading '*' from the filename key before insertion.

src/invisible_playwright/download.py

@@ -96,7 +96,9 @@ def _parse_checksums(text: str) -> dict[str, str]:
             continue
         parts = line.split()
         if len(parts) >= 2:
-            out[parts[-1]] = parts[0]
+            # sha256sum uses ' *' or '  ' prefix for binary vs text mode
+            key = parts[-1].lstrip("*")
+            out[key] = parts[0]
     return out