Task: Cut the rate instrument and leg read cost

Name the cost inside read_rate_instrument and the leg append, remove what is avoidable, and record the instruction counts before and after.

• A line-level profile names the cost of read_rate_instrument and of the leg append
• The largest avoidable cost of each is removed
• The affected tests pass and the instruction counts are recorded

Read the callees of read_rate_instrument in the callgrind tree. Remove the largest avoidable cost. Then size the leg record and decide whether the leg append is worth a change.

(Implementation strategy. Written when work starts; key decisions are distilled into the parent story's * Decisions at close, but the plan itself stays — it is the historical record of what we did.)

Every number this task claims about a tree, with the command that produced it and the commit it was measured at. A number without both is not a measurement, because a clean-up record describes a tree and trunk moves under a branch: a consumer count, a gate's component count and a suite's size all change with no code changing, and the reader of a bare number cannot tell which tree it came from. Where the raw output matters, put it in an evidence/ file beside the task and name that file in the Command column.

• Cause: a leg amount is a decimal field. own_text() formatted it as a string with to_string(), and the caller parsed it back with from_string(). The two are inverses, so the value did not change, but the format cost 6,800 instructions per call. The new record_reader::own_decimal() copies a decimal field and parses a text field. Five projections had the same helper. They all use own_decimal() now.
• The leg record is not changed. leg_data is 8,784 bytes, and its zero fill is 2.0 percent of the test. Its large members are the leg kinds. They come from the legDataType substitution group in the schema, not from an xs:choice. The column rule of the generator takes only xs:choice children, so a column for the leg kinds needs a new rule for substitution members. That is a larger change than the saving justifies. A reserve would save about 16 million instructions of the vector growth, which is 1 percent.
• If the leg zero fill matters later, extend column_members in generate_xml_bindings.py to substitution members. Leg kinds are 19.

Manual QA scenarios (scaffolded via compass add test_scenario) that verify this task. Link new ones here as they're created; the scenario doc itself links back via its "Verifies task" field.

The decimal round trip cost 6,800 instructions per leg amount. It is removed. read_rate_instrument falls from 157 to 77 million instructions, and the import test falls by 5.3 percent to 1,558,630,477. The leg record is left as it is. The ore core tests pass (378 cases, 475,410 assertions) and the benchmark gate passes.