fix: group dotted identifiers split by interleaved comments (issue712)

Identifiers may carry comments between their tokens (e.g. name -- c\n . -- c\n obj). _group only skipped whitespace when locating the tokens around a joining token, so an interleaved comment defeated the prev/next checks and the identifier was never formed.

Add an opt-in skip_comments flag to _group and enable it for group_period and group_as; the comments stay inside the resulting group. Also skip comments in get_parent_name and avoid a redundant separating space in StripCommentsFilter for the newly reachable interleaved-comment identifiers.

sqlparse/filters/others.py

@@ -63,8 +63,20 @@ class StripCommentsFilter:
             ):
                 # Insert a whitespace to ensure the following SQL produces
                 # a valid SQL (see #425).
-                if prev_ is not None and not prev_.match(T.Punctuation, '('):
-                    tlist.tokens.insert(tidx, _get_insert_token(token))
+                insert_tok = _get_insert_token(token)
+                # A plain space separator is redundant when a neighbour is
+                # already whitespace, which happens for comments interleaved
+                # inside a grouped identifier (issue712). A line break carried
+                # by the comment is always kept (see #484).
+                redundant_space = (
+                    insert_tok.is_whitespace and not insert_tok.is_newline
+                    and ((prev_ is not None and prev_.is_whitespace)
+                         or (next_ is not None and next_.is_whitespace))
+                )
+                if (prev_ is not None
+                        and not prev_.match(T.Punctuation, '(')
+                        and not redundant_space):
+                    tlist.tokens.insert(tidx, insert_tok)
                 tlist.tokens.remove(token)
                 tidx -= 1
             else: