Auto-liquidity fills a provider's free slots each run; max swap routing fee per provider

[Jossec101]

• Several swaps per auto-liquidity run. A run used to start one swap per node, so a provider's max swaps in flight was rarely reached: a Spark swap set to 5 still ran one at a time. Now a run starts as many swaps as the picked provider has free slots (its max swaps in flight minus its swaps in flight).
  • Each swap is as big as the provider allows, within what's left of the node's excess balance and, for Spark, of its wallet's room.
  • Beyond the first swap, another is added only while the run's worst-case routing fees (each swap's max routing fee) fit the swap fee budget left.
  • If a swap can't be created, the run stops and keeps the ones it started.
  • This applies to every provider. A node whose Loop max swaps in flight is above 1 will now start several Loop swaps per run.
• Max swap routing fee per provider. Loop, 40swap and Spark each have their own, set on the Nodes page next to the provider's other limits, and shown in the grid. For Spark it includes Spark's own 0.15%.
  • Auto-liquidity uses the picked provider's. So does RequestSwapOut when the caller doesn't give one.
  • The migration (AddPerProviderSwapRoutingFee) turns the node's current value into Loop's and copies it to 40swap's and Spark's, so nothing changes until you edit them.
• Tests. Unit tests for filling free slots, the slots already used, Spark's room, the excess balance, the budget cap, Loop swaps each getting their own address and routing fee, and a failed swap ending the run. The Spark E2E now starts five swaps at once with a Spark max swaps in flight of 5: all are accepted, paid in parallel by alice's LND, and exit together. No E2E drives auto-liquidity itself, but it creates swaps through the same path.