ARROW-1194: [Python] Expose MockOutputStream in pyarrow.

[robertnishihara]

This allows you to get the size of a record batch and schema through pyarrow by writing to a mock output stream. You can then use the resulting size to allocate an appropriately sized buffer to actually write to.

Example usage.

import pyarrow as pa
import pandas as pd

val = pd.DataFrame({'a': [1, 2, 3]})
record_batch = pa.RecordBatch.from_pandas(val)

# Get the size of the record batch and schema
sink = pa.MockOutputStream()
stream_writer = pa.RecordBatchStreamWriter(sink, record_batch.schema)
stream_writer.write_batch(record_batch)
size = sink.size()

[robertnishihara]

I think the Travis test failures are unrelated to this PR.

[wesm]

They're unrelated, I'm working on fixing parquet-cpp after the API changes

[wesm]

The builds should be OK now. Reviewing this

[wesm]

LGTM. Could you rebase so we hopefully get a passing build?

[robertnishihara]

Looks like the tests are passing now.

[wesm]

+1, thanks!