Fetching streams and events within a time range

[slack0]

This PR adds the long-desired feature of fetching stream data, events and raw_events by bounding them within specific start and end points, if required.

Currently, AnalysisGroup's fetch_stream(), fetch_events() and fetch_raw_events() download all the data associated with the input slug. This feature is fine with the slug corresponds to a specific device or a datablock.

There is a lot of good functionality implemented into cloud_api - persistent connection, concurrent fetching of resources, HTTP/REST error handling, fetching of API resources by object etc.

To use all of this functionality implemented into AnalysisGroup and IOTileCloudChannel for other applications - specifically arch_factory's machine models, fetch methods in AnalysisGroup need to take time range into account. This is to prevent DB loading on the cloud.

Since AnalysisGroup and its methods are used in several reports - this change could be disruptive - so enabling this by retaining the default behavior of the fetch_* methods.

What's in this PR:

• Updated fetch_stream() method to allow start and end specification. They are set to None by default. The input to these arguments could be a valid datetime-compatible string or a datetime object

• date_utils.py get_utc_ts - that accepts start, end arguments and creates a valid, ISO-format UTC timestamp from the arguments. If timezone information is not present in the arguments, then UTC timezone is forced

• Updated implementation of fetch_datapoints() in IOTileCloudChannel to account for start and end to df and data API

TODO:

• Fix test failures and add robust tests
• Update fetch_events and fetch_raw_events methods

[slack0]

@dkarchmer This is the PR for adding time range based querying in AnalysisGroup.

I am basically modifying the API calls to take in the start and end arguments - which they already support. However, I am having test failures in iotile_analytics_core. In particular, see the following call that fails. But when I use the same API on iotile.cloud (with a different stream slug), it works.

"GET /api/v1/df/?filter=s--0000-0077--0000-0000-0000-00d2--5001&format=csv&mask=1&start=2019-07-02T15%3A49%3A00Z HTTP/1.1" 200

Wanted to check with you if the test failure: https://travis-ci.org/iotile/iotile_analytics/jobs/553507199 is related / specific to the mock_cloud used in the tests.

[dkarchmer]

Seems like this has been on the back burner for long enough. The changes looks safe enough and backwards compatible to approve