Adding support for Gzip GRIB2 files.

This commit added support for reading GRIB2 files that have also been
gzipped.  Some datasets gzip GRIB2 files even though there is no benefit
in doing so.

When using the grib2io xarray backend, an error will be raised when
using Gzipped GIRB2 files.  This is a restiction within xarray.

This commit references issue #119

grib2io/_grib2io.py

@@ -111,7 +111,23 @@ def __init__(self, filename, mode='r', **kwargs):
             mode = mode+'b'
             if 'w' in mode: mode += '+'
             if 'a' in mode: mode += '+'
-        self._filehandle = builtins.open(filename,mode=mode,buffering=ONE_MB)
+
+        # Some GRIB2 files are gzipped, so check for that here, but
+        # raise error when using xarray backend.
+        if 'r' in mode:
+            self._filehandle = builtins.open(filename,mode=mode)
+            # Gzip files contain a 2-byte header b'\x1f\x8b'.
+            if self._filehandle.read(2) == b'\x1f\x8b':
+                self._filehandle.close()
+                if '_xarray_backend' in kwargs.keys():
+                    raise RuntimeError('Gzip GRIB2 files are not supported by the Xarray backend.')
+                import gzip
+                self._filehandle = gzip.open(filename,mode=mode)
+            else:
+                self._filehandle = builtins.open(filename,mode=mode,buffering=ONE_MB)
+        else:
+            self._filehandle = builtins.open(filename,mode=mode,buffering=ONE_MB)
+
         self._hasindex = False
         self._index = {}
         self.mode = mode