Sanchda/add safe copy

[sanchda]

In many configurations, process_vm_readv() may be blocked at the syscall level (e.g., common Docker seccomp profile). This provides a simple failover using writev(). Note that this mechanism is not faster than process_vm_readv().

This PR adds a few things.

• (semi-related) adds a quick failover in the common case of zero-page access
• Adds a check at static initialization on Linux for the availability of process_vm_readv()
• Implements a VmReader singleton class which holds the state for using alternative methods for reading VM. I could have put the process_vm_readv() switch here instead, but in the moment it felt right to factor this orthogonally to the current implementation
• Adds an environment variable, ECHION_ALT_VM_READ_FORCE for forcing the alternative safe reader (e.g., for benchmarking etc)

[sanchda]

Moving to draft; this code was copypasta from a different branch, but I see it may be working a little differently here. Will update when it's ready.