[Bug] [Bug] BE fails with [E-3110] parse date fail when __internal_schema.column_statistics contains '0000-00-00'

### Search before asking

- [x] I had searched in the [issues](https://github.com/apache/doris/issues?q=is%3Aissue) and found no similar issues.

### Version

4.1.0

### What's Wrong?

BE fails to initialize storage reader and continuously reports errors when `__internal_schema.column_statistics` contains illegal date strings like `'0000-00-00 00:00:00'`.

When a business table has invalid date strings (e.g. imported from legacy MySQL or skipping strict checks), the auto-analyze task collects `'0000-00-00 00:00:00'` as `min`/`max` values and writes them into `__internal_schema.column_statistics`.

When any query (or CBO optimizer) attempts to read statistics from `__internal_schema.column_statistics`, the BE Storage Reader fails to parse the string during initialization and cancels the entire query execution context with error:
`[E-3110][INVALID_ARGUMENT] parse date or datetime fail, string: '0000-00-00 00:00:00' failed to initialize storage reader.`

This creates a dead-lock where `__internal_schema.column_statistics` becomes completely unreadable (`SELECT count(1) FROM __internal_schema.column_statistics` fails), and CBO statistics loading for dependent queries gets broken.

### What You Expected?

1. Sanitization on Write: Statistics collection (ANALYZE) should sanitize or convert invalid date/datetime strings (like '0000-00-00 00:00:00') to NULL or skip them before persisting into `__internal_schema.column_statistics`.
2. Graceful Degrade on Read: BE Storage Reader should handle or ignore unparsable date values during statistics deserialization gracefully, rather than throwing a fatal error [E-3110] and canceling the entire query.

### How to Reproduce?

_No response_

### Anything Else?

- Temporary Workarounds & Mitigations:
1. Disabled global auto analyze to stop recurring failure:
`SET GLOBAL enable_auto_analyze = false;`
2. Dropped and truncated the corrupted internal statistics table to recover read ability:
`DROP STATS __internal_schema.column_statistics;`
`TRUNCATE TABLE __internal_schema.column_statistics;`

- Impact:
Because of this bug, we are forced to keep `enable_auto_analyze = false`, which prevents the cluster from automatically gathering CBO statistics for all other normal tables.

- Stack Trace snippet from BE:
0# doris::BetaRowsetReader::_init_iterator()
1# doris::BetaRowsetReader::_init_iterator_once()
2# doris::Status doris::BetaRowsetReader::_next_batch(doris::Block*)
3# doris::BetaRowsetReader::next_batch(doris::Block*)
4# doris::VCollectIterator::Level0Iterator::refresh_current_row()
5# doris::VCollectIterator::Level0Iterator::ensure_first_row_ref()
6# doris::VCollectIterator::Level1Iterator::ensure_first_row_ref()
7# doris::VCollectIterator::build_heap()
8# doris::BlockReader::_init_collect_iter()
9# doris::BlockReader::init()
10# doris::OlapScanner::_open_impl()

### Are you willing to submit PR?

- [ ] Yes I am willing to submit a PR!

### Code of Conduct

- [x] I agree to follow this project's [Code of Conduct](https://www.apache.org/foundation/policies/conduct)

Start by tracing reads of __internal_schema.column_statistics through the BE storage-reader path named in the stack trace, especially BetaRowsetReader::_init_iterator and BlockReader::init. Reproduce the failure with the reported invalid date value if possible, then inspect the ANALYZE statistics-writing path. Done means invalid date values no longer make the statistics table unreadable or cancel dependent queries.

Written by the indexing model from the issue text.