When using DHCP + clevis/tang unlocking, getting "network is unreachable" even though the network is available

Apologies in advance if this is covered elsewhere, but Google search is completely useless these days, and the various LLMs are scarcely better. Also, really appreciate all the work on this project, it's a lifesaver. When using this configuration: ```yaml network: interfaces: <mac address> dhcp: true universal: true compression: zstd modules_force_load: e1000e extra_files: busybox mount_timeout: 40s strip: true vconsole: true enable_lvm: true enable_mdraid: true ``` And booting on a Hetzner dedicated server, the network doesn't seem to be available in time for a `clevis` + `tang` unlock. Please see the attached screenshot that I pulled from a Hetzner KVM: <img width="2401" height="1720" alt="Image" src="https://github.com/user-attachments/assets/c448a3a8-861a-422c-9bfb-34f74553d68f" /> (I have redacted the various IPs involved) As shown in the screenshot, after `busybox` was escaped to, the network is clearly present and can communicate to the remote `tang` server using `wget`. I imagine the issue here is that Hetzner's DHCP response is probably quite slow. Using more or less the same configuration in QEMU, I am not able to reproduce this, as I have no idea how to "slow down" DHCP on the QEMU host. I've looked through the `booster` source code, but cannot see a way to increase the timeout. Ideally `clevis` would retry and/or be more resilient in handling potentially slow networks. I do realize that statically configuring the network interface in `booster` is probably the right way to work around this (and I will try that today), but, me being me, would love to have DHCP working in this scenario. I'd be more than willing to take a look at submitting a PR for the above, but wanted to see: a) if there was another way I was not aware of to overcome this issue, and b) if I did submit a PR for changing the way `clevis` is used if it would be accepted. Any input/feedback would be greatly appreciated. Thanks for all the work on `booster`, as I am really coming to enjoy it, and am looking forward to switching an entire fleet of servers over to using it "very soon^TM."