Use GCP bucket for temp_location when running locally (dataflow)

In `training-data-analyst/courses/machine_learning/deepdive2/end_to_end_ml/labs/preproc.ipynb`, there are two options:

- Running and storing locally (when `in_test_mode=True`; `OUTPUT_DIR = './preproc'` -- local directory)
- Running and storing on GCP (when `in_test_mode=False`; `OUTPUT_DIR='gs://{0}/babyweight/preproc/'.format(BUCKET)` -- GCS bucket)

However, when running locally, the `temp_location` needs to be a GCS bucket instead of a local directory, otherwise it will result in an error:

` "error": {
"code": 400,
"message": "Invalid extract destination URI './preproc/tmp/18bcf98db67c4b1c9a6313bc465f137f/bigquery-table-dump-*.json'. Must be a valid Google Cloud Storage path and filename/pattern.",
"errors": [
{
"message": "Invalid extract destination URI './preproc/tmp/18bcf98db67c4b1c9a6313bc465f137f/bigquery-table-dump-*.json'. Must be a valid Google Cloud Storage path and filename/pattern.",
"domain": "global",
"reason": "invalid"
}
],
"status": "INVALID_ARGUMENT"
}
`