Optimize TodoWidget filtering and alphabetical sorting allocations

[multiplex55]

Motivation

• Reduce repeated to_lowercase() allocations during todo filtering and alphabetical sorting to improve performance on large todo lists.
• Favor cheap ASCII-fast checks when possible while retaining correct Unicode-capable fallback behavior.
• Add a micro-benchmark to measure allocation/throughput improvements for large lists.

Description

• Precompute normalized filter tags once per render pass via normalized_filter_tags() and thread that slice through entry_matches_filters and filter_entries instead of lowercasing per comparison.
• Change tags_match to accept normalized filters and use eq_ignore_ascii_case before falling back to to_lowercase().contains(...) to get a fast-path for ASCII-equal matches.
• Optimize alphabetical sorting in sort_entries with a decorate-sort-undecorate approach (precompute lowercase keys, sort, then rebuild entries) to avoid repeated to_lowercase() calls inside the comparator.
• Update the filters_combine_before_sorting unit test to use the new normalized flow and add a Criterion benchmark benches/todo_widget_filtering.rs, plus register the bench in Cargo.toml.

Testing

• Ran cargo fmt which completed successfully.
• Ran cargo test --lib dashboard::widgets::todo which could not run due to the environment toolchain (rustc 1.87.0) being older than a dependency requirement (time requires rustc 1.88.0).
• Ran cargo bench --bench todo_widget_filtering --no-run which was likewise blocked by the same toolchain incompatibility, but a micro-benchmark was added and is runnable once the toolchain is updated.

---

Codex Task