__field_undef_id as name of variable with Time Series

[aclsce]

If no "name" attribute is specified, the name of variable used in Time Series file is "__field_undef_id" instead of the id of the variable.
Example :
<field id="flat" long_name="Latent heat flux" unit="W/m2" />
...
<file id="histday" timeseries="both" ts_prefix="TS_atm" >

<field field_ref="flat" level="1" />

...
gives the following file :
netcdf TS_atm_flat_20000101-20000131 {
...

float __field_undef_id_92(time_counter, lat, lon) ;

__field_undef_id_92:long_name = "Sensible heat flux" ;
__field_undef_id_92:units = "W/m2" ;
__field_undef_id_92:online_operation = "average" ;
__field_undef_id_92:interval_operation = "600 s" ;
__field_undef_id_92:interval_write = "1 d" ;
__field_undef_id_92:cell_methods = "time: mean (interval: 600 s)" ;
__field_undef_id_92:_FillValue = 9.96921e+36f ;
__field_undef_id_92:missing_value = 9.96921e+36f ;
__field_undef_id_92:coordinates = "time_centered" ;

....

[rlacroix]

Fixed in r921.