The collider problem is very complicated for platformers. I would create my own physics system based on raycasting. Sebastian Lague has a series on this on his YouTube channel. My own D and D series 2 covers this to some length. It's almost impossible to do this well in one week. You need time. There is this weird misconception in the industry that platformers are easy to make. They are not, unless you make slow ones like Limbo. Then they are somewhat easier. The cheapest solution would be a box collider, but then you encounter the problem that you cannot easily use slopes. Your game has them.

The jump height problem could be many things. My gut feeling says that your grounding is detected for less time, or something like that. Pressing the jump input at that point produces a lower jump. Another reason could be that you have your input system in FixedUpdate as opposed Update, which causes the input to be picked up more rarely, and would produce inconsistent results if you have a modular height jump.

Hard to say. Chances are that a decent platform controller may be out of your reach for now in terms of competence and confidence with code. I would recommend getting really confident and solid with less dynamic games first. I don't know your situation, so I may be misinterpreting. Point is, I've been programming for 7 years now, and I'm still worried if somebody asks me to make a platformer. :) It can go so very wrong. I've made a weird platformer before that I'm really fond of called Thrusty Ship. It's on steam and you can find the demo here on itch. You can give it a shot and see the consistency of input at play, BUT, you cannot jump in my game.

Sorry, no easy answer to your question.