feat: add comprehensive criterion benchmarks across all core crates

[mohamedmansour]

Add realistic, real-world benchmark suites for all performance-critical crates in the workspace, modeled after actual WebUI template patterns from the example apps.

Parser benchmarks (webui-parser):

• Add 4 realistic template builders: todo-app root page, shadow-DOM component, text-heavy article, and mixed-directive dashboard.
• Add parser_realistic and parser_text_vs_directive benchmark groups.
• Keep all 5 existing synthetic benchmark groups for regression tracking.

Handler benchmarks (webui-handler):

• Add handler_loop_scaling: vary array size from 10 to 2000 items.
• Add handler_condition_variety: identifier, predicate, negation, and compound AND/OR conditions with true/false/mixed state.
• Add handler_nested_components: 3-level component hierarchy (root, app, item) with 50 loop iterations.
• Add handler_state_depth: flat through 5-level nested state lookups.

Protocol benchmarks (webui-protocol):

• Add medium protocol (6-fragment todo-app equivalent).
• Add parametric large protocol generator (dashboard with N panels).
• Add protocol_size_sweep: serialize/deserialize at 5, 15, 30, 50 panel counts to measure scaling.

Expression benchmarks (webui-expressions) — NEW:

• Benchmark evaluate() across identifiers, predicates (string equality, numeric comparison, variable-vs-variable), compound AND/OR with short-circuit evaluation, negation, and realistic patterns (admin check, auth guard, item state check).

State benchmarks (webui-state) — NEW:

• Benchmark find_value_by_dotted_path across path depths (1-8 levels), value types (string/number/boolean/array/object clone cost), .length pseudo-property, missing-path early exit, large object (100 keys) lookup, and loop simulation (100-item batch lookups).

Process management refactor (xtask):

• Extract child-process lifecycle logic from dev.rs into a dedicated process.rs module with cross-platform support.
• Add graceful shutdown via SIGTERM (Unix) and CTRL_BREAK_EVENT (Windows) with a timed fallback to a forced stop.
• Spawn children in their own process groups for clean signalling.
• Replace Printer struct with direct console::style() calls, aligning with upstream terminal styling changes from chore: auto-install dev tools and unify terminal styling #63.

Xtask bench command:

• Add bench subcommand supporting parser, handler, protocol, expressions, state, and all targets.

Dependencies:

• Add criterion as a workspace dev-dependency for expressions and state.
• Add libc as a unix-only dependency for xtask (process groups).

[akroshg]

How these benchmark run? do you need to update the documentation?

[mohamedmansour]

How these benchmark run? do you need to update the documentation?

Yes, sorry, updated! You can use cargo bench -p webui-parser or you can us the task runner cargo xtask bench parser