Consider different schema representations for variable defaults (to aid highlighting variable name)

[radeksimko]

Background

Currently we pre-decode variable blocks (via earlydecoder) for type and we use that in the 2nd (real) decoding in combination with DependentBody to inform what type default is:

terraform-schema/schema/schema_merge.go

Lines 240 to 261 in faadc57

	func variableDependentBody(vars map[string]module.Variable) map[schema.SchemaKey]*schema.BodySchema {
	depBodies := make(map[schema.SchemaKey]*schema.BodySchema)
	for name, mVar := range vars {
	depKeys := schema.DependencyKeys{
	Labels: []schema.LabelDependent{
	{Index: 0, Value: name},
	},
	}
	depBodies[schema.NewSchemaKey(depKeys)] = &schema.BodySchema{
	Attributes: map[string]*schema.AttributeSchema{
	"default": {
	Expr: schema.ExprConstraints{schema.LiteralTypeExpr{Type: mVar.Type}},
	IsOptional: true,
	Description: lang.Markdown("Default value to use when variable is not explicitly set"),
	},
	},
	}
	}
	return depBodies
	}

This has the benefit of providing more contextual/relevant help for defaults, e.g.

However the downside is that we also use the existence of DependentBody to inform the user of which label or attribute has "pre-set" (enum) values, which has the less desirable effect here in the variable block, as demonstrated by semantic highlighting in hashicorp/terraform-ls#817

Proposal

Explore how else we could signify the relationship between type and default which doesn't involve DependentBody or consider adding a flag to LabelSchema which would control whether the label should be reported as dependent. We already have Completable flag there for similar reasons.