fix(assets): price the spot Market Sell instead of sending a zero

[guyverino]

What & why

Market Sell on an Assets spot row failed silently. TDoSellOrderCommand carries no
market-order flag — unlike ClosePosition, whose market_sell bool is real — so the
price=0 this path sent was never "at market". The core substituted the market's lot step
for the zero and divided the size by it: a 0.005 BTC holding reached Binance as
quantity=499.99999, was rejected five times by NOTIONAL, and nothing surfaced in the UI.

Moonbot's own Market Sell fills like a market order by pricing a LIMIT order through the
book, so this reproduces that: last * 0.8, measured from the core's own sale (ask
78528.68 → order 62822.94 → filled at 78528.67). A venue with a tighter band corrects it
itself — Bitget answered 25205 trading price cannot be below 2% and the core refilled at
its own bound.

Notable decisions

• The size stays the coin quantity. It was tempting to carry over NewOrderParams::size
  ("in the core account's balance currency"), but that documents an opening order — how
  much balance to spend — while this command says how much coin to sell. Two measured
  attempts settled it: a balance-currency notional sold 79.67 of a 198.01 AERO holding, and
  offered $0.37 of a $103 MANTRA holding, which the venue rejected for its 1 USDT minimum.
• The price is read live at confirmation, not captured at render. A wallet-derived row
  (Bitget and others expose spot holdings only through transfer_assets) carries a USDT
  valuation rather than a quote price, and a dialog left open goes stale against the book it
  has to cross.
• latest_price had to be fixed to serve that. Its fallback to the snapshot's p_last
  sat unreachable under an early return demanding history readers, which moonproto only
  builds for markets inside the client's TradeStorageScope. History is an optimisation
  here, not a requirement; the now-unreachable LatestPriceError::NoHistoryReaders is gone.
• A dropped command now says so. The confirmation previously showed nothing when no
  command was sent.

Known limitations

• Not verified live on an exchange. The defect and the two mis-sized attempts are all
  measured from core logs; the corrected build has not yet placed a real sale. The one check
  that closes it: one Market Sell on a real holding, with Sell: <whole balance> and
  SELL order DONE ! Quantity: equal to it in the core log.
• The 0.8 factor is ported from one observed core sale. Venues with a tighter band rely on
  the core's own correction, as Bitget did above.
• Errors the core reports after accepting the command still do not reach the UI — a
  rejected order is visible only in the core log. Out of scope here.

How to verify

cargo build -p moon-ui-gpui --bin moonterminal --target x86_64-pc-windows-msvc --all-targets
cargo clippy -p moon-core -p moon-ui-gpui --target x86_64-pc-windows-msvc --all-targets -- -D warnings
cargo test -p moon-core -p moon-ui-gpui --target x86_64-pc-windows-msvc

• build (all targets): exit 0
• tests: moon-core 1528 passed / 0 failed, moon-ui-gpui 1387 + 262 passed / 0 failed,
  everything else green (7 ignored, pre-existing)
• clippy: repo carries pre-existing debt, so the gate is no new finding — 16 findings in the
  touched files on this branch against 18 on origin/main, none inside this diff's hunks
  (trade.rs 163–233, read.rs 203–232)
• FireTest: not run — this change touches neither chart, render, windows nor input

🤖 Generated with Claude Code