`*usize` -> `*u32` conversion on 64-bit big-endian machine takes high
halfword of the value. As a consequence, any result returned via
`count` is unexpectedly shifted left:
u32 = 00 00 00 01 // 1
usize = 00 00 00 01 00 00 00 00 // 4294967296
Fixes following test failure:
```
$ cargo test -- tests::corpus_test
<...>
running 13 tests
memory allocation of 206158430208 bytes failed
error: test failed, to rerun pass '--lib'
```
DISCUSSION:
When compiling with `-Os` for "smallest, fastest", an error is reported in `parser.c`:
```
/Users/siegel/git/tree-sitter/lib/src/./parser.c:1368:10: error: unused variable 'did_merge' [-Werror,-Wunused-variable]
bool did_merge = ts_stack_merge(self->stack, version, previous_version_count);
^
1 error generated.
```
This is because with `NDEBUG` set, `assert(e)` collapses to `(void)0`,
which in turn means that `did_merge` does not actually get consumed.
This seems to get caught when compiling with `-Os`, but not otherwise.
Compiler version:
```
Apple clang version 13.0.0 (clang-1300.0.29.30)
Target: arm64-apple-darwin21.3.0
Thread model: posix
InstalledDir: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin
```
Due to an oversight in #1589, I added `primary_field_ids` into the
`TSLanguage` struct in a place that wasn't the end. This is not actually
backwards compatible and causes downstream failures :(
This change exposes a new `primary_state_ids` field on the `TSLanguage`
struct, and populates it by tracking the first encountered state with a
given `core_id`. (For posterity: the initial change just exposed
`core_id` and deduplicated within `ts_analyze_query`).
With this `primary_state_ids` field in place, the
`ts_query__analyze_patterns` function only needs to populate its
subgraphs with starting states that are _primary_, since non-primary
states behave identically to primary ones. This leads to large savings
across the board, since most states are not primary.
