I manage a part of the campus network and a pool of lab PCs in the large university as one of my pet projects.
We image PCs, change their default boot choice, detect unbootable computers and boot management tools all via iPXE with Dnsmasq in proxy mode. DHCP for labs is done on the router that lacks conditional statements for PXE, so we use Dnsmasq.
Dnsmasq in proxy mode would answer on the network specified in the configuration with responses that do not bear IP assignment, but have all necessary PXE options set. DHCP clients append this information to the lease responses they receive from the real DHCP server.
We run Dnsmasq on a VM in a lab cluster that at the same time provides necessary (SMB, NFS) shares to facilitate image booting and TFTP server. iSCSI boot targets are served by a SAN port on this network.
The Dnsmasq itself with TFTP for iPXE is not a resource hog, but image distribution with SMB, NFS, iSCSI is. We use 10Gbe uplinks for this because of that.
We use imaging as initial deployment step only and then manage updates, software distribution and configuration changes through Ansible roles on AWX.
You really should coordinate this with your campus IT, as our network is set up so any rogue DHCP server activity detected will trigger a port shutdown on a switch (DHCP snooping), we also do the same for STP BPDUs (BPDU guard) to make sure that students would not wreak havoc and exploit each other. At least they should be aware that you use Dnsmasq somewhere so that would not trigger unnecessary alerts.