Chart lines representing separate *tables* are linked in the special case where two grouping tags are reversed

__Steps to reproduce:__

You will need:

1. Data which undergoes first a pivot() operation to make into columns, and then a group() operation whereby it's grouped by **two** of those columns e.g.
`|> pivot(rowKey:["_time"], columnKey: ["_field"], valueColumn: "_value")`
`|> group(columns: ["_cycle","_step"], mode:"by")`
2. When viewing the **raw** output this apparently works **correctly**. There are separate tables (in my case 117), one for each *unique* combination of "_cycle" and "_step"
3. This yields the following data
[2020-12-17-09-49_chronograf_data.xlsx](https://github.com/influxdata/influxdb/files/5708644/2020-12-17-09-49_chronograf_data.xlsx)
4. The issue comes when **charting** the output

__Expected behavior:__
When charting the output I expect that lines representing each separate table of data are:
- **unconnected** (i.e. there is no line connected data from different tables)
- **unique in colour**
I also expect to have the same number of lines as I have tables. One line for each table

__Actual behavior:__
What you actually get is
- a separate colour for each **_cycle** but not for each ["_cycle","_step"] combination that the tables represent
- line joins between **certain** tables. Interestingly there are only joins between lines for tables where they have **reversed** tags. So the line for table '#'31 (_cycle 32, _step 36) is joined to the line for table '#'97 (_cycle 36, _step 32)
- this joining of similar tables means that there are significantly less lines than tables

__Environment info:__
InfluxDB version: InfluxDB Cloud 2.0

Reproduce the issue using the pivot() and group(columns: ["_cycle", "_step"], mode: "by") sequence described in the report, then compare the raw tables with the chart output. Investigate how chart series are identified and grouped for reversed tag combinations. Done means one unconnected, uniquely colored line for each table.

Written by the indexing model from the issue text.