Previously current date was used determine the offset for a given
timezone. This causes issues when the date being queried has a
different offset than today's date. Daylight savings time is an
example of that as depending on when the date is, the offset can be
different.

Fixes #7106

There are occasional test failures in nested-queries-test that is due
to the order of tests being ran and the presence (or absence of test
users. This commit just always ensures the users are there before the
test runs.

This commit differentiates between the number of columns in a
resultset and the number of columns that would actually be included in
a rendered pulse table when attempting to calculate a column
limit. Now only when the number of rendered columns (i.e. not a
description type or a details-only visualization type) exceeds the
column limit will it truncate. It will also include a CSV now if it
exceeds the column limit OR if there is a hidden column (due to it
being a description or detail-only etc).

Previously pulses and alerts that used a table visualization were
limited to 3 columns and 10 rows. This commit changes the backed to
bump that to 10 columns and 20 rows. When a resultset has more data
than will fit into a 10x20 table, it will trigger a CSV attachment
that will include the full results of the query, up to 2000 results.

This commit also limits all queries to 2000 rows, similar to how queries
issued via the UI are constrained. Users wanting more than 2000 rows
use the existing export functionality.

Specifically this occurred when filtering on a field from a nested
query. The date string wasn't getting parsed as a date and just
getting passed as a string to the database causing a type mismatch.

Fixes #6988

This commit adds query cancelling to the generic sql driver, Druid and
Presto. It cancels the future the query is running in which will
interrupt the query and allow any query cancellation logic to run. If
the database doesn't support query cancellation, or the necessary
information (i.e. query id) isn't available, the code will behave as
it did before.

This commit switches the SQL template parsing to a formally specified
EBNF grammar using instaparse. By specifying it in this way, changes
to the language become more flexible and explicit. This commit also
adds support for multiple parameters specified in an optional clause
as it's baked into the grammar.

Fixes #5492

Logical merge conflict. The code that was merged referenced a function
that was removed. Updating the test to use the new function.

BigQuery does not allown an expression in an order by clause and
instead wants name of the aliased expression from the select
clause. BQ requires this, but will break other databases (such as
Oracle). This commit will find any aggregations or aggregate
expressions and alias them in MBQL before they are convert to honeysql
structures. This allows BQ to sort by an aggregate expression.

SQLite doesn't have proper datetime support. As a result of this, the
datetimes returned from SQLite pass through the QP untouched, meaning
the datetime format is different than the other drivers. As a result,
exporting a date field to CSV from SQLite would cause a failure.

This commit gets more defensive/flexible in what kind of datetime it
will convert to a date. It will ensure what it's converting is a
string and then look for a "T" (most drivers) or a " " in the
datetime (SQLite). If it can't find either (or it's not a string) it
will pass the data through untouched.

Fixes #6944

We format date values when exported so that they don't include the
time component. This was broken if the date field had empty
values. This commit checks to ensure there's a value there before we
try to modify it.

Fixes #6901