1) Select a suitable processor.

2) Large receive receive buffers are only needed for code that are busy with other things and unable to consume received characters in the same speed as they are getting received.

3) Large transmit buffers are needed for code that don't want to hang the output routine while waiting for buffer space, and don't want to rewrite the output logic to use a state machine to continue the generation of data at a later time. So a large transmit buffer is mainly an advantage if you want simple code to be quickly able to return to the main loop to look for other things to do.

4) Value your time. Get hardware that allows you to save a number of hours of work, by letting you follow the KISS principle (Keep It Simple Stupid). For one-off hobby projects, it's generally worth it to invest in hardware and get an easier time with the programming. Looking at cents for the hardware costs is only really relevant when mass producing something.

5) Spend some time on your algorithm. The amount of state variables can vary widely with method of solving a problem. In many situations you can trade between RAM and code. A state machine can manage with very few state variables but may have a large switch statemend with lots of individual states. Or it may have a number of extra variables and intead have quite few states but with an individual state requiring lots of extra buffer space for some form of sub-state.