1.x compatibility API silently ignores "h" and "m" precision when writing

Nobody has claimed this yet.

After successfuly writing data to a 2.0.7 instance, using the 1.x compatibiliy with "m" precision I was unable to find it.

It turned out that because `models/points.go:GetPrecisionMultiplier()` silently considers anything that is not "us|ms|s" as "ns", data was written without proper multiplication of the timestamp.

I'm not sure what was the reason behind the dropping of these precisions (in 65bf4bb383693bb49e674e8cfc0a615a01388b12).
Anyway, I think that even if there is one, then either [documentation](https://docs.influxdata.com/influxdb/v2.0/reference/api/influxdb-1x/write/) should be corrected, and write handler probably made to detect this and fail,
or better, it should be supported (possibly only when using 1.x compatibility API if this is an issue for v2)

__Steps to reproduce:__
List the minimal actions needed to reproduce the behavior.

1. Write data with "m" or "h" precision and a timestamp

__Expected behavior:__
be able to get this data via a query for original timestamp

__Actual behavior:__
timestamp is incorrect.

__Environment info:__
* InfluxDB version: InfluxDB 2.0.7 (git: 2a45f0c037) build_date: 2021-06-04T19:17:40Z / latest docker 2.0 image

Start in models/points.go at GetPrecisionMultiplier(), then trace the 1.x compatibility write handler. Reproduce a timestamped write with m and h precision against the documented API behavior; done means the behavior is explicit and the original timestamp can be queried or the request fails clearly.

Written by the indexing model from the issue text.