Adjustable stop (SubmitAdjustableStop) ships without parent_id/oca_group and forces DAY -- bracket child is naked

[deepentropy]

Summary

build_order_request now reads adjusted_order_type and routes to
SubmitAdjustableStop (good). But that arm bypasses the SubmitEx extended-attrs
path, so an adjustable stop used as a bracket child ships naked, unlinked, and
DAY -- the same class as #224, for the adjustable path.

Detail

client_core.rs:1456 returns early into SubmitAdjustableStop:

if !order.adjusted_order_type.is_empty() {
    ...
    return Ok(ControlCommand::Order(OrderRequest::SubmitAdjustableStop { ... }));
}
// the SubmitEx path below (parent_id 6107, oca_group 583, tif 59) is never reached

and the FIX encoder (order_builder.rs:1042-1084) emits neither tag 6107
(parent_id) nor tag 583 (oca_group), and hard-codes tag 59 = "0" (DAY):

let mut fields: Vec<(u32, &str)> = vec![
    ..., (54, side_str), (38, &qty_str),
    (40, "3"), (99, &stop_str),
    (59, "0"),                 // tif = DAY, hard-coded
    ...                        // no 6107, no 583
    (6257, "1"), (6261, adjusted_order_type.fix_code()),
    (6258, &trigger_str), (6259, &adj_stop_str),
];

context.insert_order(Order::new(order_id, instrument, side, qty, 0, ...)) also sets
the tracked parent_id to 0.

Impact

A move-stop-to-breakeven bracket child (parent + adjustable STP child + target)
ships with the child unlinked from the parent, no OCA with the target, and DAY tif.
It rests/triggers as a standalone order rather than as a bracket leg. This is the
exact failure #224 fixed for ordinary bracket children, still present for the
adjustable-stop encode path.

Request

Route SubmitAdjustableStop through the same extended-attrs handling as SubmitEx:
carry parent_id (6107), oca_group (583), and the tif (59). Then a bracket can use an
adjustable stop child safely.