Implement $read_ndjson() and $scan_ndjson()

[etiennebacher]

TODO:

• tests
• bump news
• maybe refactor common bits with csv functions on the Rust side

[etiennebacher]

The docs failure is due to internal changes in altdoc, I'm taking care of this in #472

[etiennebacher]

Performance looks nice:

ndjson_filename <- tempfile()
df <- head(nycflights13::flights, 200000)
jsonlite::stream_out(df, file(ndjson_filename), verbose = FALSE)

bench::mark(
  ndjson   = ndjson::stream_in(ndjson_filename),
  jsonlite = jsonlite::stream_in(file(ndjson_filename), verbose = FALSE),
  jsonify  = jsonify::from_ndjson(ndjson_filename),
  arrow    = arrow::read_json_arrow(ndjson_filename),
  yyjsonr  = yyjsonr::read_ndjson_file(ndjson_filename),
  polars = polars::pl$read_ndjson(ndjson_filename),
  check = FALSE
)
#> Registered S3 method overwritten by 'jsonify':
#>   method     from    
#>   print.json jsonlite
#> Warning: Some expressions had a GC in every iteration; so filtering is
#> disabled.
#> # A tibble: 6 × 6
#>   expression      min   median `itr/sec` mem_alloc `gc/sec`
#>   <bch:expr> <bch:tm> <bch:tm>     <dbl> <bch:byt>    <dbl>
#> 1 ndjson        5.83s    5.83s     0.172   127.8MB    0.515
#> 2 jsonlite      7.31s    7.31s     0.137   278.3MB    0.547
#> 3 jsonify    746.98ms 746.98ms     1.34    105.1MB    1.34 
#> 4 arrow      393.64ms 395.87ms     2.53     27.7MB    0    
#> 5 yyjsonr    356.84ms 364.77ms     2.74     18.4MB    0    
#> 6 polars     119.52ms 126.81ms     7.88      6.9MB    0

[etiennebacher]

Thanks for the review and refactor @sorhawell, I'll do the same thing for CSV reading in another PR