The parser and StringMatchOperator class only supported "LIKE" and
"SIMILAR TO", not the inverse.  Now this is fixed.

It turns out that we never correctly implemented the INVERT flag to the
InSubqueryOperator. It was added to the toString implementation, but not
the evaluation!

Supporting code includes some SQL tests with scalar subqueries, with
some being correlated and others not.  Also, a buggy materialize-node
implementation for those who want to tinker with making IN/NOT IN
subqueries more efficient -  but the bugs will need to be fixed.

The new ANTLRv4 parser generates left-deep nested Boolean AND/OR
expressions, which need to be flattened where appropriate.  The
BooleanOperator.simplify() method takes care of this operation.

Introducing this flattening/simplification everywhere all at once seemed
to be inadvisable, so currently it is only applied in the query AST when
schemas are computed.  Additionally, only WHERE predicates, HAVING
predicates, and ON clauses are simplified/flattened.  (NATURAL JOIN and
USING join predicates are generated programmatically, not by the parser,
and are already flattened.)

The design document and information document are included under the doc
directory now.

Added a lot of documentation, particularly to the Schema class.

Made a few changes to the ArithmeticOperator and BooleanOperator
classes' simplify() methods so that we can do some basic
simplifications.

Added the StatisticsUpdater so that it's easier for students to modify
statistics based on selection predicates.

Fixed a bug in SelectClause where ORDER BY clauses were using the
FROM-clause schema rather than the SELECT-clause schema.

the `SelectClause.computeSchema` method aggregates a `resultSchema`
member by appending all of the ColumnInfos for each of the selectValues
in the statement.

However, when efforts were made to make Schema less "mutable", I think
a change was overlooked that causes all but the last selectValue's
ColumnInfos to remain in the resultSchema.

I believe this was left undetected because the resultSchema of the
SelectClause isn't really used in many places, except for nested
subqueries. This problem was found by a student, that the first out of
two columns in a nested subquery would "disappear" from the schema.

A number of bugfixes caused by some of the changes in the arithmetic
type-coercion rules, and the parser-framework changes.

COUNT(DISTINCT a) wasn't being handled correctly due to an issue in
the NanoSQLTranslator flow-control.

The SumAvg/StdDevVar aggregate implementations are now updated to
produce double-precision results when calculating their values.  The
type-converter had been changed to do integer division when the LHS
and RHS are both integers, and this was breaking tests of these
aggregate functions.

The existing SQL tests are now all assigned to the "sql" group, and a
subset are also now assigned to the "hw2" group if they should be
supported by the SimplePlanner.

The Maven pom.xml is also updated to include the "hw2" tests.

This commit includes the HW2 design document and logistics document.

It also includes the InvalidSQLException to throw when a SQL query has
a semantic error.

NanoDB client now supports shell commands of the form \cmd, which are
handled within the InteractiveClient, not by the server.  There are two
commands:

\help prints out useful help information

\source filename.sql will run all commands contained in the file
filename.sql

There are surely interesting wrinkles to iron out.  For example, if the
sourced file contains an "EXIT" command then it will exit the client.
Also, an error should probably terminate the evaluation of a sourced
file, but currently it does not.

A number of correctness tests are contained in the TestHeapTableFormat
class.  The tests are activated in the Maven test configuration.

The Buffer Manager concurrency test wasn't closing the data files it
created, so they couldn't be deleted during the cleanup phase on
Windows.  This caused test failures on Windows but not on *nix.

Somehow when upgrading the NanoDB parser to ANTLR4, I forgot to include
the SHOW STATS command.  This is essential for the storage lab, so of
course I added it back in as quickly as possible.

Some tests were failing on my box because I've got my locale set to
Portuguese; this shouldn't be an issue, since the NanoDB tests are
explicitly written in English. In order to fix Date parsing language-
agnostically, I just explicitly set the locale to `Locale.ENGLISH`.

The Buffer Manager wasn't tied into the property-registry mechanism,
so it wouldn't notice when the page-cache size was changed.  This is
now hooked together with a property-observer interface.