heatmap: reconnect to ml service with exponential backoff

Whenever we lost connection to ml service, we retry it after 1 second,
if we keep failing, we will retry with exponential backoff with max
interval as 10min.

This solves the problem that heatmap service couldn't start properly
on reboot because ml service is initialized after evdev.

Drive-by change the existng time const to be base::TimeDelta.

Bug: b:289735201
Test: Verified on Rex that heatmap palm rejection can start after reboot
Change-Id: Iafc250f91b44d7e738cf230dadbcd22d72058389
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/5369294


Reviewed-by: Rob Schonberger <robsc@chromium.org>
Commit-Queue: Jing Wang <jiwan@chromium.org>
Cr-Commit-Position: refs/heads/main@{#1273114}