The gate control logic is broken: several routines just keep it open;
other rotines close it properly; there are even other routines that
assumes that it is open without really checking or opening it.

Instead of having to manually handle the gate control and having such
troubles, let a sub-routine take care of the gate, opening it before
i2c_transfer and closing it after that. This avoids leaving the gate
into a random state.

Cc: Michael Krufky <mkrufky@linuxtv.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>