performance degradation (2x-10x) when using math builtin functions

__Proposal:__
Short summary of the feature.

__Current behavior:__
For example, when mapping over a stream of with math.sqrt and math.pow, slow down by (2x-10x). I observed that it takes an additional ~0.20s per `math` function calls. When I used x*x, instead of math.pow(), it would be about 0.20s faster.

__Desired behavior:__
At least match the performance of natively support math operations.

__Use case:__
Why is this important (helps with prioritizing requests)?
speed for high read and high cardinality computation to help with real-time data analysis. The alternative is to query raw data and compute it in code.

Disclaimer: I am new to Influx and might just not be using it right.

The issue names no files, tests, or entry points. Start by reproducing a stream mapped with math.sqrt or math.pow and compare it with equivalent native arithmetic such as x*x; investigate the performance difference with a benchmark or profiler. Done means math builtin calls no longer show the reported 2x-10x degradation.

Written by the indexing model from the issue text.