feat: add caching for timezone offsets, significantly speeds up import

this is different from pr scrapinghub#1181. that pr only makes import faster but
still incurs cost on the first usage. this one leverages an optional
cache.

closes scrapinghub#533

.gitignore

@@ -52,3 +52,4 @@ docs/_build
 
 # Other
 raw_data
+*.pkl

dateparser/timezone_parser.py

@@ -1,5 +1,6 @@
 from datetime import datetime, timedelta, timezone, tzinfo
 
+import os
 import regex as re
 
 from .timezones import timezone_info_list
@@ -85,7 +86,41 @@ def get_local_tz_offset():
 
 
 _search_regex_parts = []
-_tz_offsets = list(build_tz_offsets(_search_regex_parts))
-_search_regex = re.compile("|".join(_search_regex_parts))
-_search_regex_ignorecase = re.compile("|".join(_search_regex_parts), re.IGNORECASE)
 local_tz_offset = get_local_tz_offset()
+
+_tz_offsets = None
+_search_regex = None
+_search_regex_ignorecase = None
+
+
+def _load_offsets():
+    global _tz_offsets, _search_regex, _search_regex_ignorecase
+    _tz_offsets = list(build_tz_offsets(_search_regex_parts))
+    _search_regex = re.compile("|".join(_search_regex_parts))
+    _search_regex_ignorecase = re.compile("|".join(_search_regex_parts), re.IGNORECASE)
+
+
+if "DATEPARSER_TZ_CACHE" in os.environ:
+    import pickle
+    from pathlib import Path
+    from dateparser import __version__
+
+    path = Path(os.environ.get("DATEPARSER_TZ_CACHE_PATH", ".dateparser_tz_cache.pkl"))
+    path.parents[0].mkdir(parents=True, exist_ok=True)
+    reload = True
+
+    try:
+        with open(path, mode="rb") as file:
+            version, _tz_offsets, _search_regex, _search_regex_ignorecase = pickle.load(file)
+
+            if version == __version__:
+                reload = False
+    except Exception:
+        pass
+
+    if reload:
+        with open(path, mode="wb") as file:
+            _load_offsets()
+            pickle.dump([__version__, _tz_offsets, _search_regex, _search_regex_ignorecase], file)
+else:
+    _load_offsets()