The value that needs to be written must not be manipulated. However,
we do need to manipulate it in some case like, for example, when
configuring the offset. This is fixed by using a local variable.

The previous code was overriding the value with the manipulation
result: this is wrong and it led to the believe that this code was
working, actually it was masking another bug (fixed in another patch).

The following test in the integration test suite was passing due to
this bug, however it should have failed because the read-back value is
wrong.

    assert conf.value_get(PyAdcConf.ADC_CONF_CHN_OFFSET) == conf_rb.value_get(PyAdcConf.ADC_CONF_CHN_OFFSET)

The value in `conf` ended up to be the same as its read back `conf_rb`
due to the manilupation done direction on the pointer.

So the test needs to be fixed as well to check on this possible regression.
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>
Reported-by: Nicolas Magnin <nicolas.magnin@cern.ch>