perf: optimize shared state legend mount

[xuefei1313]

Summary

• Batch inherited shared-state scope resolution while mounting a detached graphic subtree, avoiding repeated recursive scope binding.
• Preserve custom parent-tree synchronization hooks and subsequent add/remove-child behavior.
• Compile static DiscreteLegend state definitions once per render and reuse the compiled definitions across compatible item text graphics.
• Scope the legend cache to one render and reset it on rerender/release, so it does not retain caller-owned state definitions.

Root cause

A 100-category legend creates many item groups and child graphics. During first mount, each append repeatedly resolved and synchronized the same shared-state scope; static selected/unSelected definitions were also compiled repeatedly per legend item.

Performance

For the VChart pie.small.100 case, warmed Chrome traces show a median initial render of approximately 37–38 ms, compared with 46.906 ms before the Core optimization (about 19–21% lower).

Validation

• packages/vrender-core: targeted state and batch-mount tests (29 passing).
• packages/vrender-components: DiscreteLegend tests (5 passing).
• rush compile -t @visactor/vrender-components.
• Pre-push rush test --only tag:package (all package test operations passed).