veyron/security/agent: Initial implementation using VC.
This is a bare bones implementation of an agent which stores the
identity's private key in a separate process. Currently it only
stores a single key for a single child process. Future versions will lift
these limitations.
The first attempt broke on linux because the socket addresses on Linux are
"@", which the endpoint parsing code doesn't support. This is fixed by
implementing custom names for unixfd Conn's.
Change-Id: I3122401767c8875fb88464fca2354fe201ab161d
13 files changed
