While calculating the initial frame size of the client messages,
we were adding `SIZE_OF_FRAME_LENGTH_AND_FLAGS` twice. That extra
addition is removed from the `create_initial_buffer`.

To make the `create_initial_buffer_custom` alike, I also removed
the addition from there, although there was no such problem for that
method. To make it work again, I have modified the protocol template
to perform one less subtraction of `SIZE_OF_FRAME_LENGTH_AND_FLAGS`
from the initial frame size.

Also, while I was there, I have updated the `sql_execute_codec`, and
added support for the newly added parameter.

* add a unit test that verifies the encoded requests produce the same bytearray as the Java client