Fix bug in CudaSpikeQueue creation

When the synapses_initialise_queue kernel was called with less then the
maximum of available threads per block, not all spikequeues were
initialised correctly.

brian2cuda/templates/synapses_initialise_queue.cu

@@ -17,7 +17,7 @@ namespace {
 __global__ void _run_{{codeobj_name}}_kernel(
  unsigned int _source_N,
  unsigned int _num_blocks,
- unsigned int _num_threads_per_block,
+ unsigned int _num_threads,
  double _dt,
  unsigned int _syn_N,
  unsigned int num_delays,
@@ -29,7 +29,7 @@ __global__ void _run_{{codeobj_name}}_kernel(
 
  {{pathobj}}.queue->prepare(
  tid,
- _num_threads_per_block,
+ _num_threads,
  _num_blocks,
  0,
  _source_N,
@@ -405,7 +405,7 @@ void _run_{{pathobj}}_initialise_queue()
  _run_{{codeobj_name}}_kernel<<<num_blocks, num_threads>>>(
  source_N,
  num_parallel_blocks,
- max_threads_per_block,
+ num_threads,
  dt,
  syn_N,
  num_delays,