avoid 32-bit integer overflow

change calculation to scale encoder delay and padding when sample rate
changes to use 64-bit math and avoid possible 32-bit overflow.

Bug: 67737022
Test: re-ran affected use case
Change-Id: I6e37341256f3f0e3274ffdd3bad284eaacac25ab
(cherry picked from commit 73e95a9b)
(cherry picked from commit e0d1b381)