Implement COPY … FROM STDIN queries

[ahoppen]

This implements support for COPY operations using COPY … FROM STDIN queries for fast data transfer from the client to the backend.

Performance

A quick note on local performance measurements: Inserting the numbers from 0 to 1,000,000 into a table that has two columns (INT and VARCHAR) takes ~150ms. Depending on the exact implementation, the majority of the active CPU cycles are spent converting the numbers to strings, inside string interpolation or inside ByteBuffer._setBytes. If I remove the code that sends the CopyData messages to the backend (but keep all other logic that might incur thread hopes), the test described above takes ~50ms and utilizes the CPU at ~200%, so the real bottleneck here is the Postges backend handling the data. For comparison psycopg2 takes 210ms with the data to be written already prepared in a StringIO object. So, performance-wise this PR should be good to go.

Ideas for follow-up PRs

• Check if we should buffer data sent through the PostgresCopyFromWriter to reduce the number of CopyData messages we need to send (and thus the protocol overhead). Alternatively, we can leave that kind of optimization to the client.
• Add an API that allows binary transfer of data
• Implement remaining options that can be passed to COPY FROM.
• Allow concurrently generating the data to be written and flushing a buffer to the backend.

Fixes #290

[fabianfett]

Another review round. Thanks so much for pushing through this!