bugfix: reduceSeries honors aliased names

[majedrze]

Describe the bug
Applying reduce() function after any alias* family function results in an empty result. reduceSeries derives the grouping keys from series.Tags["name"] instead of series.Name, but alias function intentionally operate on (as far as I am able to understand) series.Name.
This causes the matchers to match against the original series name, and not the aliased series name, which in turn produce an empty (or in some corner cases) straight up wrong result.

AFAIU this behaviour is not in line with graphite-web and is a correctness bug. The bug was introduced in this refactor: 0d4ebf3

CarbonAPI Version
master

Simplified query (if applicable)

assume series like:

servers.us.dc1.host01.cpu.raw_used
servers.us.dc1.host01.cpu.raw_total
servers.us.dc1.host02.cpu.raw_used
servers.us.dc1.host02.cpu.raw_total
...

use the following query:

group(
  aliasSub(aliasByNode(servers.us.dc1.host[0-9]*.cpu.raw_used,  3, 5), 'raw_used',  'cpu.actual'),
  aliasSub(aliasByNode(servers.us.dc1.host[0-9]*.cpu.raw_total, 3, 5), 'raw_total', 'cpu.max')
) | reduceSeries('asPercent', 2, 'actual', 'max')

expected result is two reduced series.
actual result is empty due to the bug.

Additional context
The patch works for me, because I am not using tags. But I am unsure on how to handle them in this function.

Should you be able to reduce on tags? Any input on this would be appreciated.

For now, I used ExtractName utility function and I think it should work in the general case for matching on metric name, but not on tags.

[teqwve]

It looks like the applyByNode function has the same issue.

[npazosmendez]

This makes sense to me

I can take care of adding a test. I am tackling another bug in this same function here #936

[npazosmendez]

Looks like some tests are failing, would you mind taking a look @majedrze ?

[majedrze]

@npazosmendez sure, tests should be clean now. I also fixed applyByNode as per @teqwve suggestion.

[npazosmendez]

Thank you! I do not see a fix for applyByNode though, did you forget to push that?
I will merge this as is now, feel free to send another PR