Even after all these decades of using calculators with an = key, after about 10 minutes using RPN, it just went "click". I totally get it!!
I like the mentality of gathering all the operands, then operating on them. In the same way I gather all the components on my desk, and only then start soldering them into the circuit.
And that's totally why RPN works. I realised very quickly that using RPN is actually the way we were taught to do things at school. You wrote down the first operand, then wrote the second one underneath it and the operation you were doing to the right of that. Then you proceeded to apply your operator to the rows and columns of digits, writing down the answer underneath.
Not an = sign in sight!
All the glory of RPN ends when you have to process large formulas with a lot of bracketed terms and you need find a way inside out and not left to right. RPN ist a way of misusing a human brain as stack memory. Humanity invented compilers to avoid that.
Algebraic notation is equally abusable. Some people seem to insist on using technology in a way that is more difficult than necessary/desirable.
But back to TE. Forth is a good option for embedded scripting languages inside TE: minimal size, complete, expandable.
Back in 1983 I spent a couple of days (only!) hacking a very simple such thing together, to peek/poke registers in some test equipment I was developing. It worked and was sufficient, but I realise that if it ever needed to be enhanced, Forth would be a better option. It made me realise that scrotty little home-grown Design Specific
Languages are almost always a mistake - and that opinion has been reinforced several times in the following years. Much better to have a Design Specific
Library in a standard language.
Nowadays MicroPython might be a sensible alternative; I believe there can even be one in the ESP8266s. I haven't investigated it, because having used makefiles and pretty-printers, I have a pathological aversion to anything where whitespace has semantic significance.
LUA might be another alternative; there is/was one for the ESP8266.