* add repo sync analytics

tracks repo metadata on first index, or on full reindex (when the
tantivy index hash is updated, for example), includes the following
info:

- repo ref
- repo provider
- size in bytes
- number of qdrant chunks
- time taken to index

sample event:

    RepoEvent {
        name: "index",
        payload: [
            ("repo_ref", String("nerdypepper/dijo")),
            ("provider", String("github")),
            ("file_count", Number(35)),
            ("chunk_count", Number(236)),
            ("bytes", String("116.59KB")),
            ("sync_time", String("4.957807429s")),
        ],
    }

---

to test this changeset, set the following values in `local_config.json`:

    {
        "analytics_key": "..",
        "analytics_data_plane": ".."
    }

run rudderstack in "live event stream" mode, and index a repository.

* track reindexs

- marks full index resets as `SchemaUpgrade`s
- marks periodic reindex jobs as `PeriodicSync`s

* clippy

* address review comments

index-stats must be consolidated before making a check for whether or
not the analytics event must be sent.

* Never start qdrant, just emit a message that it failed to run

* render bug report modal on splash screen

* Proper messaging

---------

Co-authored-by: anastasiia <[email protected]>

Co-authored-by: calyptobai <[email protected]>
Co-authored-by: calyptobai <[email protected]>
Co-authored-by: Gabriel Gordon-Hall <[email protected]>

…1143)

* Respect path filters in agent queries

* Fix autocomplete for langs

* Add directory indicator

* Add directory indicator to autocomplete

* Use central lang map instead of a separate one

* Simplify this logic

* Add position information to literals

* Add `raw_query` to exchange

* Add offsets to lang filters

* Respect the clippy

* Aggregate and rank languages from the executed queries

* Only offer lang suggestions if there's a lang filter in the query

* Partially parsed lang functions will not return results

* Allow excluding results from certain kind of results

* Normalize results somewhat

* Show lang filters more aggressively

* Add arbitrary language suggestions

* Tuning

* Make it a greedy regex

* render parsed user query in history

* add autocomplete component

* automatically convert user query between input and display

* show only file name in the autocomplete input

* use another character for autocomplete regex

* reorder query

* add content=false query param

* make the suggestions container bigger

* Apply partial lang filter

* fix and test splitUserInputAfterAutocomplete function

* improve styles for highlight background

* update autocomplete chip style

* Add langs queries to fuzzy matcher

* Autocomplete on lang even if there are no matches

* More fine tuning

* More fine tuning

* More tweaks

* Update server/bleep/src/webserver/autocomplete.rs

Co-authored-by: Gabriel Gordon-Hall <[email protected]>

* Nicer interface

* Move `raw_query` to `query`

* Fix tests

* move raw_query inside query

* Propagate certain filters to LLM

---------

Co-authored-by: rsdy <[email protected]>
Co-authored-by: rsdy <[email protected]>
Co-authored-by: Gabriel Gordon-Hall <[email protected]>