Add the functionality required to run at least one clustal job

[krs85]

The name of the branch is a misnomer... I have been working on this branch for too long (obviously), and it is very much time to rebase. I don't want to squash the commits as I think the history will prove valuable in the future. Basically what these commits do is add the functionality required to get Process$ to the point where it can cache and then skip basic programs.

Relevant issues:
#47 , #42

List of functionality added:

• Adds the functionality to skip an execution by intercepting execve and, provided it is successful and we have it cached, the next system call is changed to an exit call, effectively skipping the execution.
• If it is a failed execution, we still look it up in the cache to see if we have seen it before (and expect it to fail). Then we just let it fail.
• Add hashing functions. Add hashing of input files for the following system calls: access, read, stat, fstat, newfstat, and the open varieties. Add hashing of output files for openat, open, creat, write.
• Generate input file hashes when we first see the file access.
• Generate output file hashes at the end of the execution.
• Add serialization and deserialization functions.
• Basic cache lookup based on:
  Metadata: success or failure, executable name, env vars, cwd
  Input and output files hashes matching
• Copy output files to the cache at the end of a successful recording run.
• Overall simplification of data structures for: global cache of executions, individual execution record structs, and file access record structs.
• Extension of data structures to handle multiprocess programs and nested execution trees.
• Panic!: if trying to copy a file to the cache that's already there, if we see another output file event when this file has already been accessed as an output, if a process tries to call execve multiple times (not handling this right now, but we could just handle this as all one big execution, either all execve calls the single process does are skippable or none are), and when we expect an execve to fail (cached version does) but it succeeds.
• Current cache lookup is very conservative and is “all or nothing”. If a root process spawns many child processes, we check that ALL are skippable, and only skip if this is true.

[gatoWololo]

I'm currently seeing a compilation error because of:

[[example]]
name = "hashes_are_equal"
path = "src/hashes_are_equal.rs"

but "src/hashes_are_equal.rs" doesn't exist anymore.

Your Rust code is great and fun to read! Very understandable! I think I know how to simplify the Execution enum so that we don't need Pending root and we can avoid a bunch of the extra logic with matching on the enum variant. I can tell you about it next time we have time. I'd also be interested in implementing the code change if you like it.