perf(import): reuse auto-created groups across import chunks

lib/mv/membership/import/member_csv.ex

@@ -37,6 +37,9 @@ defmodule Mv.Membership.Import.MemberCSV do
   - `inserted` - Number of successfully created members
   - `failed` - Number of failed member creations
   - `errors` - List of `%MemberCSV.Error{}` structs (capped at 50 per import)
+  - `groups_found` - The in-memory group snapshot grown while processing this
+    chunk; thread it into the next chunk's `:groups_found` opt so groups created
+    in an earlier chunk are reused without re-reading the Group table
 
   ## Examples
 
@@ -94,7 +97,8 @@ defmodule Mv.Membership.Import.MemberCSV do
           failed: non_neg_integer(),
           errors: list(Error.t()),
           errors_truncated?: boolean(),
-          warnings: list(String.t())
+          warnings: list(String.t()),
+          groups_found: list(Mv.Membership.Group.t() | %{id: String.t(), name: String.t()})
         }
 
   alias Mv.Membership.Import.ColumnResolver
@@ -374,7 +378,7 @@ defmodule Mv.Membership.Import.MemberCSV do
       actor: actor
     }
 
-    {inserted, failed, errors, _collected_error_count, truncated?, warnings, _groups_acc} =
+    {inserted, failed, errors, _collected_error_count, truncated?, warnings, groups_acc} =
       Enum.reduce(chunk_rows_with_lines, {0, 0, [], 0, false, [], groups_found}, fn {line_number,
                                                                                      row_map},
                                                                                     {acc_inserted,
@@ -417,7 +421,8 @@ defmodule Mv.Membership.Import.MemberCSV do
        failed: failed,
        errors: Enum.reverse(errors),
        errors_truncated?: truncated?,
-       warnings: warnings
+       warnings: warnings,
+       groups_found: groups_acc
      }}
   end