Updated roadmap for traits

New roadmap for traits. Replaces #30

For v0:

• add tracking code to maintain, for each trait, the total set of caller-checked effects and unsafety for that trait (in the current package scope, among all implementations)
• add information to our audit/policy files: in addition to the public functions, we also need the traits (both implemented by and exposed by/declared by that crate). For each trait: whether caller-checked or unsafe, which specific caller-checked effects
• In the call graph in scanner, make sure trait methods are added to the call graph
• In the call graph in scanner, make sure name resolution for traits points to the abstract trait as added to the call graph.

Possibly for v0, or save for v1:

• In the call graph in scanner, add specific trait implementations in addition to the abstract trait method
• In a non-generic context when we know the trait that is pointed to, point to the specific trait implementation instead of the abstract trait method.
• For dynamic trait objects, just point to the abstract trait method

For v1:

• For dynamic trait objects, track which specific possible trait impls it could point to
• For traits called in a generic context, track which specific possible trait impls it could point to

Start by reading the scanner's call-graph handling and the audit/policy files, focusing on the unchecked v0 requirements. Done means tracking caller-checked effects and unsafety for each trait and recording the requested trait information for crates; the v1 dynamic and generic trait work is explicitly deferred.