[MRG+1] Python logging

install service_identity package in tests to prevent warnings

Relax SpiderLoader interface check

There are two functions, `configure_logging` and `log_scrapy_info` which
intend to replace scrapy.log.start and scrapy.log.scrapy_info
respectively.

Creating new functions makes evident the backward incompatible change of
using another logging system, and since the Python logging module is a
standard builtin, additional helpers make sense to be on a scrapy/utils
file.

This allows to remove `get_testlog` helper, `flushLoggedErrors` from
twisted.trial.unittest.TestCase and Twisted log observers created for
each test on conftest.py.

Changes:
 - Each module takes 'scrapy' logger and logs through it
 - Lazy string evaluation in all log messages
 - Added missing log messages in scrapy/core/engine.py
 - Contextual data such as crawler or spider instances, and failures

[MRG] Item multi inheritance fix

[MRG +1]fixed backwards compatibility for SPIDER_MANAGER_CLASS option

[MRG+1] FEED_EXPORT_FIELDS option

Some improvements for Scrapy tutorial

[MRG+1] rename SpiderManager to SpiderLoader

Remove djangoitem

Documenting Files Pipeline together with Images Pipeline

change data extraction in crawl example to be consistent with tutorial, removed statement implying mandatory usage of Item

[MRG +1]ItemLoader.load_item: iterate over copy of fields