Switch WaitGroup usage to ErrGroup

In a recent ticket I noticed that there might be cases where `EOF` errors occur, possibly due to network connectivity issues or something else, that weren't causing a full failure of the restore process so it seemed the user was generating backups in these cases that weren't fully complete as the dump process would continue after these were encountered meaning that there might have been gaps in the middle of their dumped dataset.

Switching to the ErrGroup usage here plus adding a few context error checking locations should hopefully help the dump process to stop if needed rather than provide users with a false positive.

internal/dumper/dumper.go

@@ -8,13 +8,13 @@ import (
 	"os"
 	"regexp"
 	"strings"
-	"sync"
 	"sync/atomic"
 	"time"
 
 	"github.com/planetscale/cli/internal/cmdutil"
 	"github.com/planetscale/cli/internal/printer"
 	querypb "github.com/xelabs/go-mysqlstack/sqlparser/depends/query"
+	"golang.org/x/sync/errgroup"
 
 	"go.uber.org/zap"
 )
@@ -136,8 +136,8 @@ func (d *Dumper) Run(ctx context.Context) error {
 	}
 	initPool.Put(conn)
 
-	// TODO(fatih): use errgroup
-	var wg sync.WaitGroup
+	// Adding the context here helps down below if a query issue is encountered to prevent further processing:
+	eg, egCtx := errgroup.WithContext(ctx)
 	for i, database := range databases {
 		pool, err := NewPool(d.log, d.cfg.Threads/len(databases), d.cfg.Address, d.cfg.User, d.cfg.Password, d.cfg.SessionVars, database)
 		if err != nil {
@@ -151,6 +151,11 @@ func (d *Dumper) Run(ctx context.Context) error {
 				continue
 			}
 
+			// Allows for quicker exit when using Ctrl+C at the Terminal:
+			if egCtx.Err() != nil {
+				return egCtx.Err()
+			}
+
 			conn := initPool.Get()
 			err := d.dumpTableSchema(conn, database, table, views[i])
 			if err != nil {
@@ -165,12 +170,13 @@ func (d *Dumper) Run(ctx context.Context) error {
 			}
 
 			conn = pool.Get()
-			wg.Add(1)
-			go func(conn *Connection, database string, table string) {
-				defer func() {
-					wg.Done()
-					pool.Put(conn)
-				}()
+
+			eg.Go(func() error {
+				defer pool.Put(conn)
+
+				if egCtx.Err() != nil {
+					return egCtx.Err()
+				}
 
 				d.log.Info(
 					"dumping table ...",
@@ -179,11 +185,13 @@ func (d *Dumper) Run(ctx context.Context) error {
 					zap.Int("thread_conn_id", conn.ID),
 				)
 
-				err := d.dumpTable(ctx, conn, database, table, d.cfg.UseReplica, d.cfg.UseRdonly)
+				err := d.dumpTable(egCtx, conn, database, table, d.cfg.UseReplica, d.cfg.UseRdonly)
 				if err != nil {
 					d.log.Error("error dumping table", zap.Error(err))
 				}
-			}(conn, database, table)
+
+				return nil
+			})
 		}
 	}
 
@@ -205,8 +213,13 @@ func (d *Dumper) Run(ctx context.Context) error {
 		}
 	}()
 
-	wg.Wait()
 	elapsed := time.Since(t)
+
+	if err := eg.Wait(); err != nil {
+		d.log.Error("error dumping", zap.Error(err))
+		return err
+	}
+
 	d.log.Info(
 		"dumping all done",
 		zap.Duration("elapsed_time", elapsed),