Story: Move trade-instrument parsing out of ores.qt.api

Move the trade-instrument parsing logic out of ores.qt.api and into a non-Qt component so that it can be unit-tested by linking a real library target, instead of the current stop-gap where the test target hand-enumerates the parse .cpp files and compiles them directly.

The parsing logic is already Qt-free — see Background. It lives in ores.qt.api only by historical accident. The test (ores.qt.api.tests) deliberately avoids linking ores.qt.api.lib (which pulls in Qt and needs a display), so it cherry-picks the parse source files into the test executable. That manual source list is fragile: when commit 78c68605d split swap parsing into a dedicated TU (parse_swap_instruments.cpp), the list went stale and the test failed to link with an undefined reference to ores::qt::internal::parse_swap_instrument. A stop-gap (adding the TU to the list) is in place; this story removes the root cause.

The parsing path turns a JSON envelope from the server into a concrete instrument struct and hands it to a populator interface. The relevant files in projects/ores.qt/api/ today:

Dependency audit (confirmed Qt-free): these files include only rfl/json.hpp, ores.utility/rfl/reflectors.hpp, ores.trading.api/domain/*, and ores.logging. No Q* types, no Qt headers. IInstrumentFormPopulator is a plain struct of virtual populate() overloads over trading::domain types — the Qt dialogs merely implement it.

Target home: ores.trading.api (it already owns the trading::domain::*_instrument types these functions parse into, and ores.qt.api already links ores.trading.api.lib).

projects/ores.trading/api/
├── include/ores.trading.api/
│   ├── parse/instrument_form_populator.hpp   (was IInstrumentFormPopulator.hpp)
│   └── parse/parse_trade_instrument.hpp
└── src/parse/
    ├── parse_trade_instrument.cpp
    ├── parse_swap_instruments.cpp
    └── parse_swap_impl.hpp

• IInstrumentFormPopulator, parse_trade_instrument, the swap TU, and parse_swap_impl.hpp live in ores.trading.api (non-Qt), renamespaced from ores::qt to ores::trading (or a neutral ores::trading::parse).
• ores.qt.api consumes them by linking ores.trading.api.lib; the Qt dialogs that implement the populator and call the parser are updated for the new header paths / namespace.
• ores.qt.api.tests's dispatch tests move to ores.trading.api.tests (or stay, but link ores.trading.api.lib as a normal target). The hand-maintained parse_src source list is deleted — no test enumerates individual parse .cpp files.
• The stop-gap note in ores.qt/api/tests/CMakeLists.txt is removed.
• cmake --build --preset linux-clang-debug-make and ctest pass.

• Target component is ores.trading.api, not a new component: it already owns the domain instrument types and is already a dependency of the Qt layer, so no new CMake target or dependency edge is created.
• Namespace: move from ores::qt to ores::trading (final spelling decided at implementation; ores::trading::parse is a candidate to keep the parse surface grouped).

• Changing the parsing behaviour or the wire format.
• Touching the Qt dialog UIs beyond the mechanical include/namespace updates needed to keep them compiling.
• The component regroup itself (separate story, Regroup C++ components under product-group parent directories); this story assumes the post-regroup layout (ores.trading/api/, ores.qt/api/).

• Regroup C++ components under product-group parent directories — the regroup that surfaced the stop-gap; this story is the follow-up cleanup of the parse-source enumeration.