Fix versioning. Remove objMap.

[brucen1030]

Tested with test/versioning.py
Objects in versioned buckets are not appendable.

Versioning is handled as follows.

1. In PUT (new) object, version in db table is set as follows in meta/types/object.go.

func (o *Object) GetCreateSql() (string, []interface{}, uint64) {
	version := math.MaxUint64 - uint64(o.LastModifiedTime.UnixNano())

And responsed to client with converted versionId in meta/client/tidbclient/object.go

func (t *TidbClient) PutObject(object *Object, tx DB) (err error) {
	if tx == nil {
		tx, err = t.Client.Begin()
		if err != nil {
			return err
		}
		defer func() {
			if err == nil {
				err = tx.(*sql.Tx).Commit()
			}
			if err != nil {
				tx.(*sql.Tx).Rollback()
			}
		}()
	}
	sql, args, iversion := object.GetCreateSql()
	object.VersionId = ConvertRawVersionToS3Version(iversion)

2. In ListObjectVersions, object versions will be returned in meta/client/tidbclient/bucket.go and meta/client/tidbclient/object.go.

func (t *TidbClient) GetObject(bucketName, objectName, version string) (object *Object, err error) {
	var ibucketname, iname, customattributes, acl, lastModifiedTime string
	var iversion uint64

	var row *sql.Row
	sqltext := "select bucketname,name,version,location,pool,ownerid,size,objectid,lastmodifiedtime,etag,contenttype," +
		"customattributes,acl,nullversion,deletemarker,ssetype,encryptionkey,initializationvector,type,storageclass from objects where bucketname=? and name=? "
	if version == "" {
		sqltext += "order by bucketname,name,version limit 1;"
		row = t.Client.QueryRow(sqltext, bucketName, objectName)
	} else if version == ObjectNullVersion {
		sqltext += "and nullversion=1 limit 1;" // There should be only one NullVersion object.
		row = t.Client.QueryRow(sqltext, bucketName, objectName)
	} else {
		sqltext += "and version=?;"
		internalVersion, err := ConvertS3VersionToRawVersion(version)
		if err != nil {
			return nil, ErrInternalError
		}
		row = t.Client.QueryRow(sqltext, bucketName, objectName, internalVersion)
	}
...
	object.VersionId = ConvertRawVersionToS3Version(iversion)

	return
}

3. In GET object, objectid for ceph will be get with client specified version Id.
4. In PUT objects in versioned buckets in storage/object.go, old objects will be handled in storage/object.go.

func (yig *YigStorage) checkOldObject(bucketName, objectName, versioning string) (err error) {
	switch versioning {
	case meta.VersionDisabled:
		// Delete all the bucketName + objectName.
		return yig.removeAllObjectsEntryByName(bucketName, objectName) // TODO: not deleted.

	case meta.VersionEnabled:
		// Do nothing.

	case meta.VersionSuspended:
		// Delete null version object.
		var object *meta.Object
		if object, err = yig.MetaStorage.GetObjectVersion(bucketName, objectName, meta.ObjectNullVersion, false); err != nil {
			if err == ErrNoSuchKey {
				return nil
			}
			return err
		}

		return yig.MetaStorage.DeleteObject(object, object.DeleteMarker)

	default:
		return errors.New("No Such versioning status!")
	}

	return nil
}

TODO work

1. Lifecycle and versioning adaption.
   https://docs.aws.amazon.com/AmazonS3/latest/dev/intro-lifecycle-rules.html#intro-lifecycle-rules-actions