ExecuteQuery does not stop retrying after context deadline

[zolstein]

ExecuteQuery do not stop looping immediately on context time-out. Instead, it will continue looping until some other limitation is reached. (It does not sleep, and can produce a huge number of log lines.) This is because the driver will hit the context error every time it attempts to write to the database connection, but the "context deadline exceeded" error is considered retryable. Similar issues exist for ExecuteWrite and ExecuteRead if the managed transaction returns a "context deadline exceeded" error.

#666 provides a fix, and includes a minimal test-case.