Make MODEL_TYPE fragment classification non-breaking

[llali]

Summary

Reworks the MODEL_TYPE fix so it fixes the fragment-classification bug (#176) without breaking the public API.

Background

Commit 20f9bf1 (PR #210, fixing #176) reclassified MODEL_TYPE as a TSqlFragment, but did so in a breaking way:

• Repurposed the public enum ExternalModelTypeOption { EMBEDDINGS } into a TSqlFragment class.
• Changed ExternalModelStatement.ModelType from ExternalModelTypeOption? (nullable enum) to a fragment reference.
• Renamed the enum value EMBEDDINGS → Embeddings (moved into a new ExternalModelTypeOptionKind enum).

Any downstream consumer (e.g. DacFx) referencing ExternalModelTypeOption.EMBEDDINGS or reading ModelType as an enum would fail to compile.

What this PR does instead (additive / non-breaking)

• Keeps enum ExternalModelTypeOption { EMBEDDINGS } and the ExternalModelStatement.ModelType (ExternalModelTypeOption?) member unchanged for backward compatibility.
• Adds a visitable fragment ExternalModelTypeSpecification whose OptionKind reuses the existing ExternalModelTypeOption enum, plus a new ModelTypeSpecification member on ExternalModelStatement.
• The parser populates both representations; the script generator prefers the fragment and falls back to the legacy enum.
• This resolves MODEL_TYPE not classified as TSqlFragment in CreateExternalModelStatement #176 (MODEL_TYPE is now a proper visitable fragment with real token offsets) while existing enum-based consumers keep compiling.

Files changed

• SqlScriptDom/Parser/TSql/Ast.xml
• SqlScriptDom/Parser/TSql/ExternalModelTypeOption.cs
• SqlScriptDom/Parser/TSql/TSql170.g, TSql180.g
• SqlScriptDom/ScriptDom/SqlServer/ScriptGenerator/SqlScriptGeneratorVisitor.CreateExternalModelStatement.cs
• SqlScriptDom/ScriptDom/SqlServer/ScriptGenerator/SqlScriptGenerator.AlterExternalModelStatement.cs
• Test/SqlDom/Only170SyntaxTests.cs

Testing

• Full suite: 624 passed, 0 failed on both net8.0 and net472.
• Test asserts both the new fragment (ModelTypeSpecification / OptionKind) and the legacy enum (ModelType == ExternalModelTypeOption.EMBEDDINGS) to prove backward compatibility.

[llali]

@ZEUSXXIV you commit caused breaking changes and I had to make these changes in this PR to make sure you fix stays but also they won't be any breaking changes. please let me know if you have any comment

[ZEUSXXIV]

@llali Thanks for the fix and the explanation, makes sense now. For future reference, what should I actually be checking for before raising PRs like this, so I catch breaking changes like this earlier?
Also, is the additive approach generally preferred over mutating existing types in cases like this?

[llali]

@llali Thanks for the fix and the explanation, makes sense now. For future reference, what should I actually be checking for before raising PRs like this, so I catch breaking changes like this earlier? Also, is the additive approach generally preferred over mutating existing types in cases like this?

Hi @ZEUSXXIV it's prefers not to break existing public APIs unless we can do a major version bump. I'll add more tests to prevent breaking changes like this

[ZEUSXXIV]

@llali Thanks for the fix and the explanation, makes sense now. For future reference, what should I actually be checking for before raising PRs like this, so I catch breaking changes like this earlier? Also, is the additive approach generally preferred over mutating existing types in cases like this?

Hi @ZEUSXXIV it's prefers not to break existing public APIs unless we can do a major version bump. I'll add more tests to prevent breaking changes like this

@llali Got it! Thanks again