Implementation of Builder pattern for Config

[gabhijit]

Implemented Builder for VMMConfig. (Note: This is an early
implementation, more like prototype. But almost all functionality is
present.)

Removed derive(Default) from MemoryConfig, KernelConfig and
VcpuConfig and added sensible defaults (memory = 256M, vcpu=1) (derive
default will trivially initialize this to 0 and hence we are required to
take a memory coniguration from CLI launcher).

Brought these changes to api (CLI) to demonstrate a use.

Removed required from 'memory' and 'vcpu' config.

Changed TryFrom<String> to TryFrom<&str>. Believe this is better
because it's quite possible to get &str from String than other way
round and in fact often we'll get &str from CLI parsing etc. But the
other way round is painful.

Right now passes all 'existing' tests (cargo test --all, except one
which likely fails because the right kernel image is not available to
test.)

TODO: Documentation with examples for using the Builder and Test cases
for the Builder.

[gabhijit]

A prototype pull request for the Builder pattern for config request. This does not affect the config refactoring of #55.

[andreeaflorescu]

I personally like the approach, I have a few comments on the code level, but I'll add them once other people chime in to see if we should move forward with this.

@lauralt @aghecenco @alexandruag what do you say?

The only thing that I find a bit out of hand is that in the builder we need to provide a default for KernelConfig. This could be avoided with some duplication, but we need to see if it is really important.

[lauralt]

I personally like the approach, I have a few comments on the code level, but I'll add them once other people chime in to see if we should move forward with this.

@lauralt @aghecenco @alexandruag what do you say?

I like the proposal too.

The only thing that I find a bit out of hand is that in the builder we need to provide a default for KernelConfig. This could be avoided with some duplication, but we need to see if it is really important.

Would be nice if we can avoid the Default impl for KernelConfig, and use the cmdline and himem_start default values somewhere else, but if it becomes weird we can also open an issue for it and not block this PR with that (from what it looks like we are still required to pass the path arg and a value for it, am I right?).

[andreeaflorescu]

Just a couple of nits. The code looks good to me 👍