Address Twilight's Fall improvements from issue #10

[ben-phillips]

Faction units:

• Starlancer II: repair-threshold select (any / half+ / all damaged, measured against living mechs)
• Starlancer XI: mechs join only while a ship is fielded; combat continues while ships or space-area mechs hold the space area; a "Mechs on ground" count plus a strategy select (win in space / save ground with or without sustains) decides casualty attribution and sustain use
• Colada: auto-targets the strongest unit with a capacity value
• Enigma: ignores the nebula's defender bonus (anomaly "effects", not just movement)
• A Strangled Whisper: FREE_CARGO stat — infantry and fighters ride free while the flagship lives, mechs still pay the printed capacity

Action cards:

• Feint (new): the retreat resolves at the announcement, skipping the final combat round
• Divinity: per-mode "units worth saving" gate on both save paths
• Lash: trigger list + drag-ordered destroy priority per combat mode
• Spark: Direct Hit-style target checkbox list per combat mode
• Meld: one-shot toggle melding a single own die (kernel singleDie CUSTOM_ROLL; scope select for combat vs unit-ability rolls)

Shared decks:

• Planesplitter and Valkyrie Vanguard added to the Abilities deck
• Artemiris Ascendant and Dimensional Reflection paradigms (reuse Overwing Zeta / Fragment Reality); Overwing Zeta now defaults to flagship + 2 cruisers
• Temporal Command Suite: per-genome command-token counts; Clever Genome keeps its own row (a second trigger in the same window)
• Valiant Genome: Courageous-style trigger and spend-gate lists
• Justiciar Rail: restricts only its own Space Cannon hits via a hit-pool transform, so Lightrail dock hits assign normally
• TF fighter upgrades: Fighter II model — capacity first, excess to the fleet pool (1/2 each for Hybrid Crystal Fighter)

Engine:

• TF unit-upgrade cards register ahead of the base slots so their stats settle before Capacity/Fleet Pool enforce at PREPARE
• Fighter II fleet-pool excess is measured against printed ship capacity (regardless of the enforcement toggle), and capacity enforcement no longer evicts other cargo for fleet-pool-backed overflow
• New: singleDie CUSTOM_ROLL kernel mode, SideApi.getCombatMode(), FREE_CARGO honored by both capacity and fleet pool, Infinity*0 NaN guard in capacity totals
• New invariants recorded in docs/engine-gotchas.md

[ben-phillips]

Thanks for the great feedback, I didn't understand your comment on meld until I saw how the AI decided to tackle that 🙃.

Unresolved issues: Ssruu + TCS implementation (currently Ssruu allows for only one agent behavior in combat)
Damaged starlancer XI: If some of the starlancers are damaged, it matters whether those are on the ground or in space, since those have different behavior.

[dzhiriki]

For TCS I was thinking about different approach: to have a total usage amount & checkboxes for genomes.
And implement event like GENOME_EXHAUSTED-event. It will work as it works in game: when you use genome, it triggers that event & TCS add additional use for that genome.

But, to be honest, I like your approach even more. With that one, you can check is it better to spend CCs on Genome A or save it for Genome B.

[dzhiriki]

For Colada. Current implementation, adds additional dice group, not extra die for the unit itself.
It breaks interaction with Crown of Thalnos: safe rerolls won't work.
But that opens other question: if there are 4 mechs & 4 good units (i.e. 4 Dreads). Is it better to produce a single dread with 5 dice, and 3 normal ones or is it better to have all 4 dreads roll 2 dice

[ben-phillips]

I was also realizing there's a few other cases, which I think I'll just add to this

1. Enigma should also ignore entropic scar (can sustain)
2. Lash doesn't account for the fact that not all unit costs are identical (There's a WS, mech and dread unit upgrade that change unit costs) there's also a 7 cost flagship

For the Colada question depends how much you like to gamble 😆

5.952 0 2/2/2/2, safe only
5.969 0 5/1/1/1, safe only
6.010 0 3/3/1/1, safe only
6.240 0.008 3/3/1/1, aggressive on the 3s
6.480 0.055 3/2/2/1, aggressive on all but the bare dread
6.720 0.102 2/2/2/2, aggressive everywhere

If there are multiple ships that are best damage (dreads), split the mechs between 2 of them. However, easiest to implement (all dice on best damage) is basically the same odds, so I'd lean towards implementing that, since the logic is very straightforward.

[dzhiriki]

Nice catches. I know the main game pretty well, but haven't played TF, so most of those interactions are unknown to me. I was just trying to find them from ability texts. Probably there are a bunch of other ones that we didn't think :)
Yeah, for Colada it is a very niche case, so it can be implemented in a simple way.

[ben-phillips]

Colada changes should now be applied (got reverted in a merge conflict). I forgot that crown adds +1 on re-rolls, but it seems like it's still best to have it on the highest damage ship.