Balance accuracy stats of T3 Arty

[lL1l1]

Issue

Related Discord thread. On top of the DPS and cost issues mentioned in #6481, there are also significant issues in the effectiveness of artillery due to inaccuracy vs splash damage. T3 Artillery often comes out at the stage of the game where there are only a few highly defended key targets to attack, such as other artillery, game enders, ACUs, nukes, and SMD. This means that high splash, low accuracy, fast reload artillery is generally much less favored than the opposite which is better at breaking through shields in one point and destroying a small target. On the other hand, high splash radius with a high enough accuracy begins multiplies its damage by hitting multiple shields at the same time, which is also unfortunately countered by microing shields so only one shield is active at a time, in which case DPS becomes the most important stat (balanced in the other PR, although it assumes perfect accuracy, it is close enough).

Description of the proposed changes

Builds on top of the changes in #6481.

Balance of arty accuracy:

All faction arties are spawned with full adjacency and in the same exact spot. They shoot at the exact same invincible shields which measure how much direct DPS (no overspill) the arties are dealing to the shields overall. This results in a % DPS multiplier for the arty since it can hit multiple shields. The firing randomness is changed iteratively until the % DPS multipliers match for all arties, which in theory means that they perform equally well against multiple shields despite their accuracy and damage radius differences.

Testing done on the proposed changes

See the testprocedures.txt file in the commits.

Checklist

- [ ] Changes are annotated, including comments where useful

• Changes are documented in the changelog for the next game version

[Tagada14]

The mass cost of Emissary is reversed in the balance change snippet

[lL1l1]

I don't really have a solid test plan for testing artillery randomness, but this is better than nothing so I would try it out in games.

I guess a proper way is to create some test unit that generates a target base template and then records data about DPS/time so that you could run the sim at +15 for a bit to get the right numbers, but it seems excessive and then isn't even good at modeling player shield micro, where it is better for arty to be a little OP vs bases than underpowered.