Allow a default dialect at the semantic model level

## Problem

Most models use the same dialect for nearly every field and metric, so this gets repeated throughout the file:

```yaml
expression:
dialects:
- dialect: ANSI_SQL
expression: SUM(orders.amount)
```

The repetition makes otherwise simple models harder to read and maintain.

## Proposal

Allow a semantic model to declare a default dialect and use a string shorthand for expressions:

```yaml
semantic_model:
- name: sales
default_dialect: ANSI_SQL
datasets:
- name: orders
source: analytics.orders
fields:
- name: amount
expression: amount
metrics:
- name: revenue
expression: SUM(orders.amount)
```

Explicit dialects would still be supported when an expression needs an override:

```yaml
expression:
dialects:
- dialect: ANSI_SQL
expression: DATE_TRUNC('month', orders.created_at)
- dialect: SNOWFLAKE
expression: DATE_TRUNC('MONTH', orders.created_at)
```

Existing files would remain valid unchanged. The shorthand would only be allowed when `default_dialect` is present.

This is related to #52, but takes a smaller, backward-compatible approach: it keeps per-expression dialect overrides and scopes the default to each semantic model.

The issue names no files, tests, or entry points. Start by locating the semantic-model schema and validation paths for default_dialect and both expression forms. Done means models with a default accept string expressions, explicit dialect overrides still work, and existing files remain valid.

Written by the indexing model from the issue text.