If you insist on getting your animations specifically from Koikatsu, it will be quite the tedious process. I will also drop a quick warning that the end result will be a lot of smaller animations due to how Koikatsu lets you change motions and speed on the fly. So 1 motion in Koikatsu will end up as a good 20-30 animations in Unity that you will have to stitch together yourself. I was about to write out an entire guide on how to go about it but I found a pretty good video that shows off one of the ways you can go about extracting animations here: https://www.youtube.com/watch?v=Ezsy6kwgBE0. Some of the comments under the video also cover the Blender > Unity conversion. If you have any specific questions, I will gladly answer them.

As for how you can set up your scene, there's 2 ways of going about it. You can either have it all on an avatar, or you can make a world and play it out there. Using an avatar would likely be the easier way of going about it if you've never messed with worlds before, but the end result of a world gives way more freedom.

If you're putting it on an avatar, you could either put one of the animations F/M on your avatar and then the second one on a secondary one that you have right under your avatar root, or you could have 2 separate avatars under your root with each their own animation, male and female. For the separate avatar(s), you just give them their own animator and put the animation in. When you toggle the avatar on, it will automatically play the animation from 0 every time. You can also go to your animation clip and click on "Loop" in the inspector to have it loop endlessly as long as it's active. You would likely have to go into play mode in Unity at first and mess around with their placements so they line up with each other. Alternatively, you can give them both a way of getting placed in the world. Modular avatar offers an avatar component you can put on your avatar that does exactly that, so instead of the avatar following your root around, it sticks to wherever in the world you decide to freeze it.

If you're putting it in a world, you could do the same as an avatar and have 2 avatars in the world just playing animations. Could add toggles to toggle them individually if you want. You could set up their animators so they offer multiple animations each that you can cycle between. You could even go and make yourself a simple MMD world where you can have as many different animations and avatar combos as you'd like. I've made an MMD world for this kind of stuff in the past, and it is definitely the way I would recommend the most if you're willing to put the work in. One thing to note when it comes to a world, however, is that they do not support SPS; you would need to use DPS. You can use SPS as a penetrator into DPS, but DPS cannot be used into SPS, so depending on the setup you're looking for here, you might need to set up different systems on your avatars.

This was a long write-up, and I might have skimmed over and/or missed something, so if you have any questions about anything, feel free to ask.

I still feel a bit stuck when it comes to how to either make Milltina-like models work in other programs (Koikatsu, mmd) or keeping it in vrchat and importing animation files instead. It seems sticking to vrchat might be easier, but I'm not sure how I could have more than one body/character in a sex scene of my making if I intend to be alone but want a male and female.

As for MMD; while importing models into Koikatsu is difficult with animations being easy, it is the complete opposite for MMD. Importing your model into MMD is as simple as putting the avatar into Blender and then exporting it as a PMX instead of FBX. All of the MMD shapekeys (allowing expressions in mmd) should already be set up without you having to do anything as most avatar base creators make them compatible before selling. If there are clothing items you want to have on your avatar when importing into MMD, you'd first have to merge said clothing in Blender before exporting the full thing.
As for the animations; animating in MMD is (in my opinion) very slow and tedious, finding motions that you can play with your avatar online is also quite difficult. I've found a fair amount myself but it's required a lot of searching around in the right places.

If you're new to Blender and/or Koikatsu along with MMD, I would highly suggest you try to extract animations from Koikatsu as a starter. Alternatively just find animations already made for Unity online, then take those animations and put them on your avatar. There are plenty of guides online on how to do everything I've mentioned here so getting it to work should be very possible if you take your time to learn it all.

Technically you could animate it and play it out in Unity or straight up in VRChat with said animation too, though if you're taking that route it's highly suggested to use Blender for making the animations.

I'm just interested in making my own videos and pictures for personal enjoyment casually.