[bugfix](load) fix coredump in ordinal index flush (apache#9518)

commit apache#9123 introduce the bug. bitshuffle page return error when
page is full, so scalar column write cannot switch to next page, which make
ordinal index is null when flush.

All page builder should return ok when page full, and column writer procedure
shoud be append_data, check is_page_full, switch to next page

Co-authored-by: yixiutt <yixiu@selectdb.com>

be/src/olap/rowset/segment_v2/binary_dict_page.cpp

@@ -75,13 +75,13 @@ Status BinaryDictPageBuilder::add(const uint8_t* vals, size_t* count) {
  }
 
  for (int i = 0; i < *count; ++i, ++src) {
+ if (is_page_full()) {
+ break;
+ }
  auto iter = _dictionary.find(*src);
  if (iter != _dictionary.end()) {
  value_code = iter->second;
  } else {
- if (_dict_builder->is_page_full()) {
- break;
- }
  Slice dict_item(src->data, src->size);
  if (src->size > 0) {
  char* item_mem = (char*)_pool.allocate(src->size);

be/src/olap/rowset/segment_v2/bitshuffle_page.h

@@ -105,7 +105,7 @@ class BitshufflePageBuilder : public PageBuilder {
  DCHECK(!_finished);
  if (_remain_element_capacity <= 0) {
  *count = 0;
- return Status::RuntimeError("page is full.");
+ return Status::OK();
  }
  int to_add = std::min<int>(_remain_element_capacity, *count);
  int to_add_size = to_add * SIZE_OF_TYPE;

be/src/olap/rowset/segment_v2/page_builder.h

@@ -49,7 +49,9 @@ class PageBuilder {
  // Add a sequence of values to the page.
  // The number of values actually added will be returned through count, which may be less
  // than requested if the page is full.
- //
+
+ // check page if full before truly add, return ok when page is full so that column write
+ // will switch to next page
  // vals size should be decided according to the page build type
  // TODO make sure vals is naturally-aligned to its type so that impls can use aligned load
  // instead of memcpy to copy values.