In Python 3 dividing by int will call obj.__truediv__(). This operator
was missing from F-expressions.

Also changed several occurrences of 'request' to 'response'.

This is more idiomatic and avoids returning a list on Python 2 and
an iterator on Python 3.

A test compared list directly against a range(). This is py3
incompatible. Fixed by using list(range()).

An ordering test had two problems related to dict.items() usage:
  - It assumed the order of the dict was non-randomized
  - It indexed to the dict.items() which is py3 incompatible.

I fixed the test by using dict['rank'] directly, where rank is the
column tested on the values queryset.

when computing the length of a dictionary. This fails on Python 3.

In Python 3, HTMLParser does not support bytestrings.

Perfomed some style cleanup while I was in the area.

This reverts commit b109ff80 and
complement test cases. The change was too hasty, as some form
values cannot be json-serialized as is.

Refs #18340

Thanks Andrews Medina for the initial patch.

Fixed a regression introduced in 59a65598.

Deferred models now automatically update only the fields which are
loaded from the db (with .only() or .defer()). In addition, any field
set manually after the load is updated on save.

This patch removes an unconditional float(value) conversion from db
backend default convert_values() method. This can cause problems when
aggregating over character fields for example. In addition, Oracle
and SQLite already return the bare value from their convert_values().

In the long term the converting should be done by fields, and the
fields should then call database backend specific converters when
needed. The current setup is inflexible for 3rd party fields.

Thanks to Merlijn van Deen for the original patch.

Thanks Aymeric Augustin for the review.

The outer query's set of reusable joins (can_reuse) was passed to the
inner query's add_filter call. This was incorrect.