Turn math cache into its own object

[xmariachi]

Purpose of Changes and their Description

Turning math cache into an object instead of using global vars.

Are these changes tested and documented?

• If tested, please describe how. If not, why tests are not needed. -- Added extensive unit tests, in both the cache and the usage.
• If documented, please describe where. If not, describe why docs are not needed. -- It is a performance change, not a functionality change, so it should not need its own documentation.
• Added to Unreleased section of CHANGELOG.md?

[spooktheducks]

Per comments in private messages with @jmdkastro and @xmariachi , I say we hold on approvals here pending:

• solid benchmarking showing that this cache is actually necessary (as opposed to just causing code bloat)
• a more performantly constructed key for the cache map (we can reduce Decimal into hashable bytes much more efficiently than converting to string with fmt.Sprintf)
• an understanding as to why we're fragmenting the cache, which somewhat negates the effects of caching math in the first place