The user directory doesn't seem to be tokenizing user IDs entirely correctly.

This issue has been migrated from [#7590](https://github.com/matrix-org/synapse/issues/7590).

---

### Description

Looking at the contents of my homeserver's `user_directory_search` table, I'm seeing the tokenization of some user ID parts to be incomplete:

```
synapse=# select * from user_directory_search limit 5;
user_id | vector
-------------------------------+---------------------------------------------
@negritoofcourse:matrix.org | 'matrix.org':2 'negritoofcours':1A,3B
@freenode_jin1200:matrix.org | 'freenod':1A 'jin1200':2A,4B 'matrix.org':3
@freenode_Xenguy__:matrix.org | 'freenod':1A 'matrix.org':3 'xenguy':2A,4B
@morethanabitoff:matrix.org | 'matrix.org':2 'morethanabitoff':1A,3B
@curtisthe:matrix.org | 'curtisth':1A,3B 'matrix.org':2
(5 rows)
```

For instance, we see `@curtisthe:matrix.org` has `curtisth` without the `e` as a token, same with `@negritoofcourse:matrix.org` and the various `freenod`s.

It looks like there might be an off-by-one error somewhere in the `user_directory_background_update` process.

Doesn't seem to be a huge impact as people haven't reported unusable results, but probably a good thing to fix.

Synapse v1.14.0.