speech: Fix behaviour with audio in ROM

In case the audio is received from VSI through
the host machine, the target has to wait some
time for the new buffer to be received which
allow some time for the connectivity task to
send and receive data to and from the cloud.

In case the audio is located inside the ROM,
the target does not have to wait as the full
audio clip is available. Hence, keeps running
without allowing some time for the
connectivity task to send and receive
data to and from the cloud, to fix this
issue a delay of 50 ticks was added
based on testing the behaviour of the
FVP and FPGA on the three available endpoints.

Signed-off-by: Ahmed.Ismail <Ahmed.Ismail@arm.com>