Add error handling to object.read_stream for reading corrupted text files from GCS

[EmersonYe]

pydatalab/google/datalab/storage/_object.py

Lines 190 to 205 in d903190

	def read_stream(self, start_offset=0, byte_count=None):
	"""Reads the content of this object as text.
	Args:
	start_offset: the start offset of bytes to read.
	byte_count: the number of bytes to read. If None, it reads to the end.
	Returns:
	The text content within the object.
	Raises:
	Exception if there was an error requesting the object's content.
	"""
	try:
	return self._api.object_download(self._bucket, self._key,
	start_offset=start_offset, byte_count=byte_count)
	except Exception as e:
	raise e

If a text file in GCS has any non-ascii characters, attempting to read_stream the file results in the following error message:
UnicodeDecodeError: 'ascii' codec can't decode byte 0x8e in position 54628: ordinal not in range(128)

I suggest adding an errors argument, like in Python 3's built-in open function. The option to ignore encoding errors or replace malformed data would make reading text files from GCS in DataLab much easier. The workaround I resorted to was to download the text locally and clean it before re-uploading to GCS

Additionally, the download and read_lines functions are not documented on the included storage documentation (http://localhost:8081/notebooks/datalab/docs/tutorials/Storage/Storage%20APIs.ipynb)