fix: make tan and atan2 compatible

[kazuyukitanimura]

Which issue does this PR close?

Closes: #1897

Rationale for this change

#1897 claims tan is incompatible; however, what is really incompatible is atan2 that is failing in the same test in #1896

The correct results are

 atan2(-0.0, -0.0) =  -π    <= Comet answer
 atan2(-0.0, +0.0) = -0.0
 atan2(+0.0, -0.0) = +π
 atan2(+0.0, +0.0) = +0.0   <= Spark answer

Looks like Spark adds +0.0 to inputs in order to convert atan2(-0.0, -0.0) to atan2(+0.0, +0.0)

What changes are included in this PR?

Fixed atan2 and enabled tan

How are these changes tested?

[andygrove]

Thanks @kazuyukitanimura. Could you add SQL file based tests, since this is preferred approach now. See #3854 for example.

[kazuyukitanimura]

Thanks @kazuyukitanimura. Could you add SQL file based tests, since this is preferred approach now. See #3854 for example.

Thanks @andygrove, updated spark/src/test/resources/sql-tests/expressions/math/atan2.sql

[comphead]

if it is only 1 input value which returns incorrect response, should we just hardcode the response?

[kazuyukitanimura]

Can done separately. I haven't decided how we should handle this. Because this is more like Spark side issue.

[kazuyukitanimura]

Actually we need do more tests for atan2 to enable e.g.

 atan2(-0.0, -0.0) 
 atan2(-0.0, +0.0)
 atan2(+0.0, -0.0)
 atan2(+0.0, +0.0)
 atan2(-0.0, -1.0) 
 atan2(-1.0, -0.0)
 atan2(-0.0, +1.0)
 atan2(+1.0, -0.0)
...

[kazuyukitanimura]

@comphead I ended up fixing atan2 as well. Please take a look