Add CoreDHCP to Helm

[rainest]

Remove dnsmasq from the Helm chart. Add Deployments for atftpd and CoreDHCP.

Related to #86

[alexlovelltroy]

@rainest This is still marked as WIP. Do you anticipate trying to close it this month?

[rainest]

@alexlovelltroy end of month may be iffy, as nothing will happen after the 23rd. Depends on how much we're okay with a less than ideal implementation.

The current set of commits brings it in line with the updated coredhcp with built-in TFTP, though that appears a bit broken still--my attempts to get files that should be available resulted in timeouts, so I need to figure out what's up with the routing there.

The initContainer population of the the TFTP artifacts is probably functional, but poor design and quite slow. IDK that there's any particularly good option there without image volumes, which won't really be an option for a while. PV initialization would deal with restart slowness, but IIRC Helm hook behavior makes that very annoying to deal with.

IMO that storage is probably better handled out of band, but that doesn't really work for an all-in-one quickstart.

Nevermind, moot point as of c379d16. These are now baked in.

Aside from the missing hosts file, I can probably live with the ugly templated YAML doc within a ConfigMap for now.

[rainest]

332d053 is ready-ish with the lazy templated YAML string within a ConfigMap nonsense. It can't actually serve TFTP outside the cluster without OpenCHAMI/coresmd#14, but won't really change after that--just a coresmd version bump and the additional plugin toggle.

TFTP pulls from within the cluster do work fine, so aside from the CNI fun, it's working as expected.

#86 (comment) explains this. dnsmasq may have a single-port mode similar to coresmd#14 below, but I don't see that we had it configured, so I don't think this was likely working before.

Haven't tried DHCP; need to find a simple way to test that.

[rainest]

With coredhcp 0.3.0 this can successfully test TFTP pull. The test is cluster-internal and thus doesn't quite test the single-port NAT traversal, though I've tried that externally successfully.