You're overcomplicating a relatively simple situation.
If you're going to keep all clients on one subnet, you DO NOT connect the two routers Lan port to WAN port This method is ONLY used when you want two separate subnets...........period.
Yes, you probably could connect from the LAN port of the upstream router (router 1) to the WAN port of the downstream router (router 2) and keep them in a single subnet. But, this complicates matters a whole lot and in the end, it comes down to one simple question:
"Why would you want to make more work for yourself and overcomplicate something relatively simple by doing it this way?"
We're professionals. We're telling you, in a single subnet scenario you ALWAYS use the aforementioned "LAN port to LAN port" type connection.
Now to address your two (actually 3) questions:
1) In this small an environment there is no "cost" to routing. Routing happens so fast you cannot notice the time taken to open the packet and route it one whole hop by any human sense. As to having one LAN port used up, spend $60.00 USD and buy an 8 port, 1000 Mbps switch and connect it to another LAN port and suddenly, like magic, you now have 10 LAN ports available
If the WAN (internet) port isn't being used then no, you need not configure anything on it with regard to type of internet connection.
NOTE: I've seen some SOHO Routers that have a setting that allows you to use the WAN port as a LAN port. If yours is capable of that, then you could enable that feature, plug router 1 into the WAN port on router 2 (with it acting as a LAN port) and then still have all 4 LAN ports available for clients AND, not have to route
2) Correct, you would only use this method if you're setting up a second, separate subnet.
I would like to confirm that with the second option, it is possible to use both routers with DHCP activated. I think it is.
Since both are on separate subnets, there could be no conflict. Technically speaking you can have two separate DHCP servers running on the same subnet. I've had occasion to do this which I will explain below. The main reason for not doing it though is to avoid problems. With a single DHCP server running in your network it keeps things nice and simple. (If you're not familiar with the KISS principle, google it. We all try our best to live with it in the computing world)
I once setup a training lab using two SOHO Routers in a remote location from our main buildings. Since there were 30 laptops in this training lab, I didn't want to overload a single SOHO router and that's why I had two. I grabbed a block of 30 IP's in the particular subnet we were using for that location and divided it into two. I enabled DHCP on the both routers and assigned half the IP's to one, and half to the other. Something like:
DHCP Scope = 192.168.1.10 to 192.168.1.25
DHCP Scope = 192.168.1.26 to 192.168.1.40
You could do this in any LAN to LAN setup if you really wanted to. But, it adds complexity (again, always apply the KISS principle whenever possible) and makes it possible to screw up your DHCP and have neither one work.
I did what I did above knowing once the 15 IP's for router 1 were used, nobody else could connect to it and would therefore connect to router 2. This kept both from being overloaded.
It matters not how straight the gate,
How charged with punishments the scroll,
I am the master of my fate;
I am the captain of my soul.