power: qpnp-fg-gen4: Fix ESR fast calibration calculations

ESR_DELTA is having 17 bit encoding instead of 12 bit encoding
which other ESR parameters have. So, account in that while doing
the calculations. Also, apply the delta without filtering for the
first iteration of ESR fast calibration so that ESR calculation
can converge faster.

Change-Id: I5aae1c53a5eb29c219f31c3faf98c1872f753074
Signed-off-by: Subbaraman Narayanamurthy <subbaram@codeaurora.org>