Almanac: Retry requests to the Almanac server on network failure

With App Preload Service rolling out to production, we know that
requests to the server have about a 95% success rate. About 2% of errors
are caused by either NETWORK_CHANGED or NAME_NOT_RESOLVED, which can be
easily retried by SimpleURLLoader.

Retrying these requests should cause the overall APS success rate to
improve, we can monitor UMA to determine whether this is the case. I've
made the change to Promise Apps as well, since this also queries Almanac
without a retry config.

Bug: None
Change-Id: Id1637f3ebfb34ff01b769d489212fd873515b146
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/4680761
Commit-Queue: Tim Sergeant <tsergeant@chromium.org>
Reviewed-by: Melissa Zhang <melzhang@chromium.org>
Reviewed-by: Vivian Pao <vpao@google.com>
Cr-Commit-Position: refs/heads/main@{#1169137}