Split customer-facing calls from technology source

[lockefox]

pandas-datareader splits their API calls from the specific platform (Yahoo, Google, Quandl, etc). Expand paths like datareader/coins/ to datareader/coins/hitbtc/ and datareader/coins/cryptocompare/

Full paths may not be required. But make private platform-specific libraries for separating concerns.

[lockefox]

Expect API pathing something like:

import prosper.warehouse.coins as coins

coins.get_orderbook_hitbtc('BTC', 'asks')

but allow get_orderbook_hitbtc() to source deeper:

"""prosper.datareader.coins.__init__.py?"""
from . import hitbtc
from . import cryptocompare

def get_orderbook_hitbtc(coin, which_book, currency, logger=DEFAULT_LOGGER):
    """get the current orderbook from hitbtc"""
    coin_info = hitbtc.get_coin_info(coin)
    ...

Move all the __all__ scoped functions to be easier to reach, and platform-weirdness segregated by provider rather than type

[lockefox]

v2 refactor branch: https://github.com/EVEprosper/ProsperDatareader/tree/v2-refactor

[lockefox]

released in v2.0.0