Allow specifying a max fee for static Loop In to force splitting across multiple channels to the last hop

**Is your feature request related to a problem? Please describe.**
When using static Loop In (`loop static in ... --last_hop `) with more than one channel open to the same last-hop peer, there's currently no way to force the payment to split across those channels. LND's pathfinding only attempts to shard a payment when no single route/channel can carry the full HTLC — if one channel happens to be big enough on its own, LND will always prefer the single-HTLC route, even if the user would rather spread liquidity across both channels.

This matters for users actively managing inbound liquidity across multiple channels to the same peer (e.g. Loop's own routing node) — right now the only way to force a split is to manually shrink the swap amount below the capacity of any single channel, which isn't practical for larger swaps or automated tooling.

**Describe the solution you'd like**

Add a max-fee argument to the loop-in request (e.g. `--max_fee_sat` / `--max_fee_ppm` at the routing/HTLC level, distinct from the existing `--max_swap_fee_sat` / `--max_swap_fee_ppm` which caps Loop's *service* fee).

Setting a low-enough max routing fee would eliminate the single-HTLC route from consideration (a single larger HTLC generally costs more in routing fees than an equivalent smaller shard), forcing the payment to split across the available channels as a side effect — without needing to build new explicit splitting/shard-count logic.

An alternative would be an explicit shard-count argument on the loop-in request, but the max-fee approach seems simpler since it reuses a lever that already exists conceptually in LND's payment API rather than requiring new splitting logic.

(Note for maintainers: in LND terms this means the payment falls back to its existing MPP shard-splitting behavior once the single-HTLC route is priced out of consideration.)

**Describe alternatives you've considered**

MPP?

**Additional context**
Static Loop In user with 2 channels open to the Loop peer, wanting a large loop-in swap (static or normal) to route inbound liquidity across both channels rather than concentrating it on whichever single channel happens to have enough capacity.

Start at the static Loop In command and request handling, tracing the existing max swap fee arguments into the payment-routing call and the LND payment API. Compare the proposed routing fee limits with existing MPP behavior; done means a distinct routing-fee limit is accepted for Loop In and verified to prevent an otherwise eligible single-HTLC route when the limit requires splitting.

Written by the indexing model from the issue text.