This will solve the issue that abort the connection transaction when we
skip the tests.

This implements the support to encode/decode JSON
data to/from database and creating columns of type
JSON using a native type [1] supported by PostgreSQL
from version 9.2.

[1] http://www.postgresql.org/docs/9.2/static/datatype-json.html

Reason since MySQL 5.6.6-m9 the `sql_mode` default value is
`NO_ENGINE_SUBSTITUTION`.
This default parameter change is out of control from Rails.

This test verifies Rails not overriding the default `@@GLOBAL.sql_mode` value
by checking if `@@GLOBAL.sql_mode` is the same as `@@SESSION.sql_mode`.

Didn't fail the test because adapter#query happens to
not call raw connection's #query, but don't want to count
on that and have a fragile test.

Full test requiring manual intervention was fine, but
w/ simulated disconnect, assertion was backward & still
passing. Was several kinds of wrong.

It doesn't serve much purpose now that ActiveRecord::Base.all returns a
Relation.

The code is moved to active_record_deprecated_finders.

Previously it returned an Array.

If you want an array, call e.g. `Post.to_a` rather than `Post.all`. This
is more explicit.

In most cases this should not break existing code, since
Relations use method_missing to delegate unknown methods to #to_a
anyway.

Conflicts:
	activerecord/test/cases/query_cache_test.rb

This commit needs to be reverted because it introduces difficulties when
using sqlite3 in development and other databases in production. This
happens because when you create time column in sqlite3, it's dumped as
datetime in schema.rb file.

This reverts commit 57d534ee, reversing
changes made to 20f049fb.

Conflicts:

	activerecord/test/cases/adapters/sqlite3/sqlite3_adapter_test.rb

This patch addresses the difficulty of retrieving datetime fields. By default, the database holds a higher precision than the time as a String.

This issue is discussed at length at the following links:
- [#3519](https://github.com/rails/rails/issues/3519)
- [#3520](https://github.com/rails/rails/issues/3520)

Also, kudos to @mattscilipoti

Improve the derivation of HABTM join table name to take account of nesting.
It now takes the table names of the two models, sorts them lexically and
then joins them, stripping any common prefix from the second table name.

Some examples:

  Top level models
  (Category <=> Product)
  Old: categories_products
  New: categories_products

  Top level models with a global table_name_prefix
  (Category <=> Product)
  Old: site_categories_products
  New: site_categories_products

  Nested models in a module without a table_name_prefix method
  (Admin::Category <=> Admin::Product)
  Old: categories_products
  New: categories_products

  Nested models in a module with a table_name_prefix method
  (Admin::Category <=> Admin::Product)
  Old: categories_products
  New: admin_categories_products

  Nested models in a parent model
  (Catalog::Category <=> Catalog::Product)
  Old: categories_products
  New: catalog_categories_products

  Nested models in different parent models
  (Catalog::Category <=> Content::Page)
  Old: categories_pages
  New: catalog_categories_content_pages

Also as part of this commit the validity checks for HABTM assocations have
been moved to ActiveRecord::Reflection One side effect of this is to move when
the exceptions are raised from the point of declaration to when the association
is built. This is consistant with other association validity checks.

escape_hstore uses quotation marks around keys and values only if it
seems necessary. However, it currently breaks in the presence of some
non-ASCII characters. Instead of trying to guess exactly which
characters are safe, it seems better to always use quotes.