When retrieving data from S3, account for chunking

[horsto]

There seems to be an issue with retrieval of larger ("external") files from S3 compatible buckets. I documented the error here: #1083. I brought it up as issue in the minio-api repository as well: minio/minio-py#1280.

The problem is that files can be returned as chunks via the minio API, such that .data cannot be called directly. Instead, .stream and subsequent concatenation of bytes seems to work for all cases. This behavior is already correctly implemented in fget() (

datajoint-python/datajoint/s3.py

Line 88 in ee95f02

for d in data.stream(1 << 16):

)

[horsto]

This PR changes the .data method to .stream in the get() method for S3 (external) objects, which accommodates chunking of larger files (opposed to assuming that every / the whole file is loaded at once).

[dimitri-yatsenko]

@horsto Would you merge this PR horsto#1 ?

[horsto]

Done!