Show active filters as removable chips with a full reset link

[loevgaard]

Resolves #51

What

Applied filters now render as chips above the search results, matching the mockup in the issue:

• Clicking a chip's ✕ removes that single filter (via the existing AJAX swap + pushState flow; the chips are plain links, so they degrade to full page loads without JavaScript).
• A Reset all filters link clears every filter while keeping the query (q), the sorting (s), and any unrelated query parameters.
• The chips also render on the no-results page, where the filter form is not rendered at all today — so users can un-filter their way back to results instead of being stranded.
• Removing a filter always resets the page parameter (p).

How

• New ActiveFiltersProvider (src/Provider/ActiveFilters/) computes the chips server side from the request's f query parameters. Filter types are resolved through the index metadata (the same source the form/filter builders use), and the value guards mirror ArrayFilterBuilder/BooleanFilterBuilder/FloatFilterBuilder, so a chip is only shown for a filter the engine would actually apply. Unknown facet names are skipped.
• Templates get the chips via a new ssm_active_filters() Twig function (SearchRuntime) and a new search/_active_filters.html.twig partial, included inside #search-form in _results.html.twig and _no_results.html.twig so it is re-rendered on every AJAX swap.
• search.js gains a shared #navigate(url) (extracted from #submitForm), a public navigate(), click delegation for the chip links (wired before the form check so it also works on the no-results <div>), new onFilterRemove/onFiltersReset options, and new search:filter-removed/search:filters-reset events. All additive.
• Labels: choice chips show the raw facet value (already localized at index time); boolean chips use new count-free active_filters.facet.* keys with a humanized fallback for untranslated custom facets; range chips read e.g. "Price: from 50". New translation keys added to all 12 locales.

Behavior change: disjunctive facetStats

The range inputs are prefilled with the facet's bounds and submitted with every form interaction, so the URL always carries f[price][min|max] even when the user never touched them — naively that would render a phantom "Price" chip after any interaction. A range chip is therefore only shown when a bound actually narrows the facet's stats.

For that comparison to work, SearchEngine::provideSearchResult() now merges facetStats across the disjunctive per-facet queries the same way it already merges facetDistribution (previously the stats came only from the fully-filtered main query, against which a user-set bound never registers as narrowing). Side effect: range inputs now prefill with bounds that exclude their own filter, which is the standard disjunctive faceting behavior.

Notes for extenders

• Integrators overriding _results.html.twig/_no_results.html.twig keep their current markup and can opt in with one include line; opting out with stock templates is overriding _active_filters.html.twig with an empty file.
• ActiveFiltersProviderInterface is aliased, so chip computation can be decorated/replaced.
• Known limitation: a hand-crafted falsy boolean filter (f[onSale]=0) filters to false but gets no chip — that state is unreachable through the shipped UI.

Testing

• Unit: SearchEngineTest (disjunctive stats merge) and ActiveFiltersProviderTest (15 cases: choice/bool/range chips, narrowing suppression, invalid values, unknown facets, reset URL, taxon-style paths).
• Functional: ActiveFiltersProviderTest proves the wiring + real Meilisearch stats end to end, including the "prefilled bounds produce no chip" case.
• E2E: new "active filters" describe block in search.spec.ts (6 tests) covering chip rendering, removal, the phantom-price-chip guard, the no-results rescue, and reset keeping query + sorting.

All suites pass locally: 66 unit, 13 functional, 20 e2e; PHPStan (level max), ECS, Rector, lint:container, lint:twig, composer validate --strict and composer normalize --dry-run are clean.

[codecov]

Codecov Report

❌ Patch coverage is 85.58559% with 16 lines in your changes missing coverage. Please review.
✅ Project coverage is 41.11%. Comparing base (14bef06) to head (43f3a43).
⚠️ Report is 2 commits behind head on master.

Files with missing lines	Patch %	Lines
.../Provider/ActiveFilters/ActiveFilterCollection.php	0.00%	7 Missing ⚠️
src/Twig/SearchRuntime.php	0.00%	6 Missing ⚠️
...c/Provider/ActiveFilters/ActiveFiltersProvider.php	97.87%	2 Missing ⚠️
src/Provider/ActiveFilters/ActiveFilter.php	0.00%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##             master     #144      +/-   ##
============================================
+ Coverage     37.92%   41.11%   +3.18%     
- Complexity      711      768      +57     
============================================
  Files           131      135       +4     
  Lines          2286     2403     +117     
============================================
+ Hits            867      988     +121     
+ Misses         1419     1415       -4     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.