Add exponential backoff for failed requests

[yatharthranjan]

TO ensure that something like #62 does not happen in the future and to better handle consistent failure of requests (and avoid being banned by Garmin's systems), we should add support for exponential backoff at multiple levels:

1. User level: If all the routes for a user fail, we should keep backing off exponentially for that user.
2. Route level: If a route is failing for majority of the users, we should backoff exponentially for that route.

[yatharthranjan]

Some optimisations for the specific case in #62 are already added in #70

[yatharthranjan]

Another optimisation we can do is keep a cache of users with 401 still pushing the data. This way we can reply with 200 for users already removed or outdated in our systme back to garmin to reduce the errors reported on garmin portal. If possible we should fix #45 first