feat: implement structured pattern matching with nested constructor support

[metalurgical]

Introduce a full pattern matching pipeline for match expressions, including typed pattern representation, validation, and a structured decision tree.

This establishes a foundation for deterministic and type-safe match evaluation, enabling correct handling of nested patterns and complex destructuring.

Additionally extract tests for parse to new file to reduce file bloat and improve maintainability.

Overview Diagram

flowchart TD
    A["match syntax"] --> B{"Old vs New"}

    B -->|Old| C["2 arms only → normalize"]
    B -->|New| D["multiple arms → analyze → validate → compile"]

    C --> E["Match.left/right"]
    D --> E

Loading

---

Expanded Diagram

flowchart TD
    A["match syntax"] --> B{"Old vs New"}

    B -->|Old| C1["exactly 2 arms required"]
    C1 --> C2["classify arm pair"]
    C2 --> C3["Left/Right or None/Some or False/True"]
    C3 --> E["Match.left/right"]

    B -->|New| D1["parse multiple arms"]
    D1 --> D2["convert to structured pattern representation"]
    D2 --> D3["assign types to patterns"]

    D3 --> D4["ensure all arms use the same match kind"]
    D4 --> D5["check for duplicate arms"]
    D5 --> D6["check for overlapping patterns"]
    D6 --> D7["check for unreachable arms"]
    D7 --> D8["check exhaustiveness"]

    D8 --> D9["build decision tree"]
    D9 --> D10["convert back to two-arm form"]
    D10 --> E["Match.left/right"]

Loading

[metalurgical]

Resolves #250
Closes #250

[KyrylR]

Quite tough to review, as a single commit contains too many changes

The main reason for having smaller commits is to isolate possible regressions, so when the commit history follows the rule of a "logical sequence of clearly defined independent changes" it is easier to debug

For example, if at some point a bug/regression is discovered, I can write a test and find the exact commit that introduced it
But if this regression happened in a commit like this, it would be much harder to say where exactly

This is why I have to ask you to split this MR into smaller commits