fix(fulltext): stop BMW scoring iterators on a stale posting block (#…

…3496)

### What

Fix a server abort in fulltext search with many query terms. Fixes
#3495.

`match_text` with more than 15 terms could terminate the process from a
query thread:

```
TerminateHandler: Unhandled Exception: number of bits is unsupported
```

(or `Decode posting FAILEDF@src/storage/common/int_encoder.cppm:109`),
thrown under `BlockMaxWandIterator::Next` ->
`TermDocIterator::BM25Score` ->
`SkipIndexDecoder::DecodeCurrentTFBuffer`.

### Approach

- `BlockMaxWandIterator::Next` only partially sorted `sorted_iterators_`
for more than `SORT_SKIP_THRESHOLD` terms. With the pivot in the
unsorted tail, iterators in `[0, pivot]` could sit on an earlier doc;
BMW called `NextShallow(d)` on them (block cursor moves, nothing
decoded) and then `BM25Score()` without a `Next(d)`. The partial sort is
removed, the iterators are always fully sorted.
- `PostingIterator::SkipTo` now invalidates `current_row_id_` when the
block cursor moves, so `GetCurrentTF` / `GetCurrentTTF` /
`GetCurrentDocPayload` return 0 instead of decoding from the reader's
stale position. Before, the only guard was an `assert` in
`MultiPostingDecoder::DecodeCurrentTFBuffer`, compiled out in release
builds.
- `FragmentTask::OnExecute` catches `std::exception` and fails the
query. FastPFor throws `std::logic_error`, which nothing caught before.

Update version to 0.7.3 (#3413)

### Summary

Fix CI process (#3406)

Fix 'No UV'

Signed-off-by: Jin Hai <haijin.chn@gmail.com>

Update version to 0.7.1 (#3390)

### What problem does this PR solve?

As title

### Type of change

- [x] Python SDK impacted, Need to update PyPI
- [x] Other (please describe): update version

Co-authored-by: Jin Hai <haijin.chn@gmail.com>

Fix: Multiple Memory Indexers Cause BM25 Calculation Error (#3340)

Close #3339

### Summary

`ColumnIndexReader` may contain multiple memory indexers (one per
segment). However, `FullTextColumnLengthReader::GetDocTermCount()` was
only using a single `memory_indexer_` variable, causing `total_df_ <
doc_freq_` crash during BM25 calculation.

### Reproduction Steps

1. Set `optimize_interval` to 3600s
2. Run `TestMultipleIndexTypesImport` with snapshot restore enabled
3. The test will crash with `total_df_ < doc_freq_` error

### Expected behavior

_No response_

### Additional information

`UnrecoverableError(fmt::format("total_df_ {} is less than doc_freq_
{}", total_df_, doc_freq_));`

Fix PhysicalFilter::Execute (#3342)

### What problem does this PR solve?

When table is empty, prev_op_state->data_block_array_.empty() is true in
PhysicalFilter, we should return empty result instead of crash.

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)

Fix release (#3343)

### What problem does this PR solve?

Correct docker image name.

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)

Signed-off-by: noob <yixiao121314@outlook.com>