Fix synchronization bug in bool parquet mukernels (rapidsai#17302)

This fixes a synchronization bug in the parquet microkernels for plain-decoding bools.  This closes [several](NVIDIA/spark-rapids#11715) timing [issues](NVIDIA/spark-rapids#11716) found during testing of spark-rapids.

Authors:
  - Paul Mattione (https://github.com/pmattione-nvidia)

Approvers:
  - Bradley Dice (https://github.com/bdice)
  - Vukasin Milovanovic (https://github.com/vuule)

URL: rapidsai#17302

cpp/src/io/parquet/decode_fixed.cu

@@ -848,6 +848,7 @@ inline __device__ void bool_plain_decode(page_state_s* s, state_buf* sb, int t,
 {
   int pos              = s->dict_pos;
   int const target_pos = pos + to_decode;
+  __syncthreads();  // Make sure all threads have read dict_pos before it changes at the end.
 
   while (pos < target_pos) {
     int const batch_len = min(target_pos - pos, decode_block_size_t);