Unify name/wildcard/regex lookup and miss-behavior across holders

[gupichon]

Description

Name-based lookup (__getitem__, .get(), find_elements()) was implemented independently in at least six places across ElementHolder and the holder/array hierarchy, each with its own idea of what counts as a wildcard, whether re: regex is supported, and what happens on a miss (None, an exception, AttributeError, or a silent empty result). This surfaced while adding ToolHolder (#375): sr.live.bpm["BPM_001"] raised TypeError (no __getitem__ on singular holders) while sr.live["BPM_001"] and sr.live.bpms["BPM_001"] both "worked" with disagreeing semantics. This PR defines one shared name/wildcard resolution convention (pyaml/common/name_matching.py) and applies it consistently to every holder and array.

Related Issue

• Fixes Feature: Unify name/wildcard access and miss-behavior across all holders #433

Features/issues described there are:

• new feature: add __getitem__ to GenericElementHolder, RFHolder, RFTransmitterHolder, ToolHolder, and DiagnosticHolder — implemented by routing all of them, plus the existing ElementHolder.__getitem__/find_elements() and ElementArray.__getitem__/GenericArrayHolder.__getitem__, through one shared resolver (resolve_names()/is_wildcard()) instead of each reimplementing its own rules.
• bugfix: fnmatch.fnmatch() normalizes through os.path.normcase(), so the same wildcard pattern matched case-insensitively on Windows and case-sensitively on Linux CI — switched every call site to fnmatch.fnmatchcase().
• bugfix: an invalid re:-prefixed regex used to leak a raw re.error; it now raises PyAMLException like every other lookup failure.
• decision, not code: GenericArrayHolder.__getitem__ (e.g. magnets["QF1*"]) keeps searching individual element names, not registered array/family names — this is Feature: Complete magnet holder navigation #373's recent, deliberately tested behavior. Documented explicitly instead of unifying it with .get(name)'s separate array/family-name namespace, since unifying would revert Feature: Complete magnet holder navigation #373.
• scope addition beyond the issue text: also added __getitem__ to DiagnosticHolder and RFTransmitterHolder, which have the exact same shape as RFHolder/ToolHolder and didn't exist yet when the issue was written (they landed with Feature: Add a public diagnostic holder #372 and Feature: Add a public tuning and measurement tool holder #375).
• scope addition, from thread discussion: regex (re: prefix) is kept, reversing what the issue currently says, and extended everywhere the resolver is used, not just find_elements(). A list/tuple of patterns (literal, wildcard, or regex, mixed) is also now accepted everywhere a single pattern is, unioned and de-duplicated.

Changes to existing functionality

• ElementHolder.__getitem__: a literal exact name with no match now raises PyAMLException instead of returning None. Breaking change — any caller relying on the old None-on-miss must now catch PyAMLException or check membership first.
• ElementArray.__getitem__ / GenericArrayHolder.__getitem__: same literal-miss fix (raises instead of silently returning an empty array); wildcard trigger now also includes [, not just */?; case-sensitive matching via fnmatchcase; gains re: and list-of-patterns support. A wildcard/regex pattern matching nothing is unaffected, still an empty collection, not an error.
• ElementHolder.find_elements(): reimplemented as a thin wrapper around the shared resolver, same public behavior plus the [-trigger, case-sensitivity, and error-wrapping fixes above, and now also accepts a list of patterns directly.

Testing

The following tests (compatible with pytest) were added:

• tests/common/test_name_matching.py: the shared resolver directly — literal hit/miss, wildcard hit/empty, regex hit/empty/invalid, case sensitivity, list-of-patterns (union/dedup/order, one missing literal raises)
• tests/common/test_element_holder_collection.py: ElementHolder.__getitem__ / find_elements() literal-miss-raises, [-as-wildcard, re:, list-of-patterns, case sensitivity, plus new tests combining wildcard/regex/list selections with &/|/- set operations (non-trivial intersection/union/difference, and the empty-result-stays-a-plain-list quirk under the new selection styles)
• tests/arrays/test_array_selection_types.py: ElementArray.__getitem__ literal-miss-raises, bracket-only wildcard, regex, list-of-patterns, case sensitivity
• tests/common/test_array_holder_navigation.py: GenericElementHolder.__getitem__ via .magnet, .combined_function_magnet, .serialized_magnet; a realistic example reproducing the QForTest YAML exclusion family via [] selection and -
• tests/diagnostics/test_diagnostic_accessors.py: DiagnosticHolder.__getitem__ and .diagnostic.bpm[...]
• tests/rf/test_rf_holder.py (new): RFHolder/RFTransmitterHolder.__getitem__
• tests/tuning_tools/test_tool_holder.py (new): ToolHolder.__getitem__

Verify that your checklist complies with the project

• New and existing unit tests pass locally (pytest tests -n auto: 389 passed, 6 skipped; one unrelated pre-existing flake, test_tuning_orm, is a Windows temp-file PermissionError unconnected to this change)
• Tests were added to prove that all features/changes are effective
• The code is commented where appropriate
• Any existing features are not broken, aside from the explicit, documented breaking change above (literal-miss now raises instead of returning None)