fix(dbt): Support syncing derived metrics on legacy dbt versions

[Vitor-Avila]

PR#254 introduced a new pattern for mapping metrics:

metric_map = {metric["unique_id"]: metric for metric in metrics}

This ended up affecting our ability to parse derived metrics, given the metric macro renders the metric name, rather than its unique_id. As a consequence, this condition was never met (since token.sql() would return the macro name):

for token in tokens:
      if token.sql() in metrics:

This section of the code is only used by legacy dbt versions -- for new versions we're relying on MetricFlow to provide the metric syntax.

This PR re-introduces the old mapping structure specifically for this context:

metric_map = {metric["name"]: metric for metric in metrics}

It also includes the dialect as part of the metric schema, to assist sqlglot when parsing the syntax. I was facing an issue when trying to parse below metric syntax since it's BQ-specific:

SAFE_DIVIDE(
   SUM(
      IF(
         `product_line` = "Classic Cars",
         price_each * 0.80,
         price_each * 0.70
      )
   ),
   revenue_verbose_name_from_dbt
)

[betodealmeida]

Ah, great catch!