Feat/3316 add mermaid diagram export to schema

[heaven00]

Description

Adding capability to generate mermaid diagrams from schema, the representation in mermaid includes the following

• column names
• data type
• primary key or foreign key
• relationships ( first_entity, cardinality, second_entity, label)

Related Issues

• feat: dlt.Schema.to_mermaid() #3316

Additional Context

Task List

• Create a module dlt/helpers/mermaid.py
• Build the diagram by manipulating strings without additional Python dependencies
• Write modular functions and tests for generating columns, tables, etc.
• Add dlt.Schema.to_mermaid() method
• Add to the CLI dlt schema --format mermaid FILE and dlt pipeline NAME schema --format mermaid
• Update CLI docs
• Add example export to the docs

[heaven00]

@zilto I seem to have steered away from regular parameters a bit in the implementation.

• Do I need to work with TStoredSchema instead of the Schema object?
• Is it okay to include table_names as parameter or stick to show everything and just provide removing dlt private tables?

[zilto]

This PR is almost there! There are few things to align with existing to_dot() and to_dbml() for consistency and some unit tests to add

[zilto]

I believe it's worth having a _to_mermaid_column() function that generates a single column. Then we have:

• _to_mermaid_column()
• _to_mermaid_table()
• _to_mermaid_reference()
• schema_to_mermaid()

Then, add unit tests to this function to check if it produces the expected results for:

• no hint
• each individual hint
• 2+ hints

[zilto]

Additional arg

I think it would be useful to have a top-level arg hide_columns: bool (available on schema_to_mermaid() and dlt.Schema.to_mermaid(). if hide_columns is True then we just have table names and references. This gives a much more compact visual when you have a lot of tables (this could be added in .to_dot() and .to_dbml())

[heaven00]

hey @zilto almost done with the feedback you provided (thanks for that)

one question, at the moment when we include all the dlt tables the _dlt_version renders separately

though the version is populated by every table should it be any different? things pending

• hide_columns feature
• cli integration and testing with the new parameters for pipeline and schema
• documentation update

[zilto]

closing in favor of #3364