TSQL: Remove dependency on ANSI keyword lists

[jpers36]

Up to this point, the TSQL dialect has incorporated the ANSI dialect's keyword list into its own keyword list. The upside of this has been the ability to rely on ANSI's definitions even when those definitions include references to keywords not existent in TSQL. The downside is that some non-keywords may flag as keywords when parsing TSQL.

This PR aims to rip off the bandage. This is driven by an issue where I have a stored procedure that can't parse because the parser incorrectly flags ROWS as a reserved keyword. So this PR clears the dialect keyword list before applying TSQL's list, and incorporates all changes needed to make that work.

TSQL keyword changes:

• Added a number of unreserved keywords that are actually used by TSQL but were previously not in the list

TSQL dialect changes:

• Clear ANSI keywords from list
• Add BaseExpressionElementGrammar to override ANSI and remove IntervalExpressionSegment reference
• Add IsClauseGrammar to override ANSI and remove NAN reference
• Add LikeGrammar to override ANSI and remove RLIKE/ILIKE
• Add FunctionContentsGrammar to override ANSI and remove SEPARATOR
• Remove IntervalExpressionSegment reference from Expression_D_Grammar
• remove CreateExtensionStatementSegment, CreateModelStatementSegment, DropModelStatementSegment, and DescribeStatementSegment from StatementSegment
• Add CreateSequenceOptionsSegment to properly represent TSQL functionality
• Adjust ColumnConstraintSegment to remove invalid AUTO_INCREMENT and UNSIGNED options, and remove reference to invalid CommentClauseSegment
• Remove CreateExtensionStatementSegment, CreateModelStatementSement, DropModelStatementSegment, OverlapsClauseSegment Nothing() stubs, replacing them with removals elsewhere
• Add MLTableExpressionSegment Nothing() stub to remove ML keyword; add TO DO to possibly represent PREDICT statment here
• Rename PartitionByClause to PartitionClauseSegment to match ANSI dialect naming convention; fix type name to match as well
• Add AlterTableSegment to override ANSI and remove MODIFY
• Add SetOperatorSegment to override ANSI and remove MINUS
• Add WindowSpecificationSegment to override ANSI and remove the unsupported functionality to name a window; this needs to be done because PARTITION's status change from a reserved keyword to an unreserved keyword otherwise confuses this definition

Test cases:

• Added TSQL non-keywords to select test case
• Added more detailed case to sequence test case
• Added ALTER table test cases
• Minor adjustments due to the PartitionByClause/PartitionClauseSegment change

[codecov]

Codecov Report

Merging #2170 (da96bcf) into main (f20bfae) will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##              main     #2170   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files          148       148           
  Lines        10796     10798    +2     
=========================================
+ Hits         10796     10798    +2     

Impacted Files	Coverage Δ
src/sqlfluff/dialects/dialect_tsql_keywords.py	100.00% <ø> (ø)
src/sqlfluff/dialects/dialect_tsql.py	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f20bfae...da96bcf. Read the comment docs.

[jpy-git]

@jpers36 looks great, couple points to clear up 👍

[jpy-git]

Rather than have a stub I think you can just remove this from the StatementSegment like this:

[jpers36]

@jpy-git thanks for the syntax help! I moved a couple more stubs to this design now that I know how.

[jpy-git]

@jpers36 really great stuff, thanks for driving the T-SQL syntax coverage forwards 💪 LTGM 🚀 ⭐