42522: colexec: add unordered distinct r=yuzefovich a=yuzefovich

**distsql: add a test comparing distinct operator and processor**

Release note: None

**colexec: minor cleanup of hashjoiner_tmpl**

This commit removes some extra empty lines from the generated code by
appending `-` to the end of the template definitions. Also, it refactors
one conditional into an assignment.

Release note: None

**colexec: fix a bug in hash table when null equality is allowed**

Previously we incorrectly were handling the case of allowed null
equality in the hash table (which is used by the hash grouper and
unordered distinct) - we were resetting groupID of the tuple on the
probe side of that tuple has NULL value in the equality column whereas
the build tuple has a non-NULL value. This could result in incorrectly
populated `same` array and is now fixed.

Release note: None

**colexec: add unordered distinct**

This commit adds an unordered distinct operator. The operator simply
builds a hashTable on the first call to Next() on the whole input, then
iterates over all of the tuples to check whether the tuple is the "head"
of a linked list that contain all of the tuples that are equal on
distinct columns. Only the "head" is included into the big selection
vector. Once the big selection vector is populated, the operator
proceeds to returning the batches according to a chunk of the selection
vector.

This doesn't take into account whether the input is partially ordered,
and this item is left as a follow-up work.

Fixes: #39240.

Release note (sql change): vectorized execution engine now supports
DISTINCT on unordered input.

**colexec: fix hash grouper**

Previously, hash grouper could hit an internal error of index outside of
range when it was looking for the "head" of a linked list. Now this is
fixed by refactoring the code in question. I think the code is now more
comprehensible.

There is no release note since hash grouper so far has been used only
with `experimental_on`.

Release note: None

**distsql: enhance testing columnar aggregator against processor**

This commit adds an ability to insert NULL values into grouping columns
and increases the number of grouping columns tested (previously it was
hard-coded as one). ANY_NOT_NULL aggregate is now skipped since it
returns non-deterministic output.

This change required modifications to the test harness to allow the
results of aggregation to deviate slightly when operating on float
values because direct string matching will not work in some cases.

Also, this commit extracts out arguments to the test harness into
a separate struct.

Release note: None

43288: ui: network latency page r=dhartunian a=elkmaster

created new menu item for network latencies reports, redesign, added sort-by and filtering features

Resolves: #43206

Release note (ui): none

![Screenshot 2019-12-18 at 16 43 02](https://user-images.githubusercontent.com/12850886/71095590-8b6a0600-21b5-11ea-89ee-314869e2a455.png)
![Screenshot 2019-12-18 at 16 43 15](https://user-images.githubusercontent.com/12850886/71095591-8b6a0600-21b5-11ea-991c-28202f4c5de4.png)


44365: roachtest: adjust tpchvec test r=yuzefovich a=yuzefovich

This commit adds the print out of all vec ON and vec OFF times if the
test is failed because of slow vec ON performance.

Also, in order to reduce noise, we now require that the best vec ON time
is worse than the best vec OFF time. This addresses a case when a query
has two fast runtimes and two slow ones (I don't know why it happens,
but sometimes the first two runs of a query, especially with a small
order number like Q3, will be slow with vec ON).

Fixes: #44331.

Release note: None

Co-authored-by: Yahor Yuzefovich <yahor@cockroachlabs.com>
Co-authored-by: Vlad <carrott9@gmail.com>
Co-authored-by: Vlad Los <carrott9@gmail.com>