Add definition of zerocopy

[dfrg]

This adds a new Definitions section to the README along with a proposed definition for zerocopy. I believe this accurately represents the result of the relevant discussion from the font-types PR and the associated issue.

[dfrg]

I think the discussion here is interesting and useful but not particularly germane to the PR. I propose to narrow the definition to:

zerocopy

The direct reading/writing of scalars and, wherever possible, aggregates (structs and slices) through reinterpretation of pointers to raw bytes in memory.

… and move the implementation strategy discussion to an issue so we can continue there.

[rsheeter]

+1 to the refined definition, I think that's a much more direct assertion of "zerocopy means X"

EDIT: I'm +1 to merge with that version. If we want to capture the graph traversal aspect that's fine but maybe in a subsequent PR.