Drop DOM parser

[tindzk]

According to the browser benchmarks, the internal parser is twice as
fast as the DOM-based parser:

Benchmark: Parse HTML w/o attributes
- depth=2:
  units: 7
  iterations: 8801
  run time: 59092 μs/it ± 2280
- depth=6:
  units: 127
  iterations: 830
  run time: 854302 μs/it ± 251892
- depth=10:
  units: 2047
  iterations: 65
  run time: 8001891 μs/it ± 309583
- depth=14:
  units: 32767
  iterations: 4
  run time: 140500000 μs/it ± 5250000

Summary:
  Unit growth: 18.1x, 16.1x, 16.0x
  Run time growth: 14.5x, 9.4x, 17.6x

Benchmark: Parse HTML w/ attributes via DOM
- depth=2:
  units: 7
  iterations: 3148
  run time: 142371 μs/it ± 16459
- depth=6:
  units: 127
  iterations: 482
  run time: 1057465 μs/it ± 20120
- depth=10:
  units: 2047
  iterations: 31
  run time: 16211693 μs/it ± 336693
- depth=14:
  units: 32767
  iterations: 2
  run time: 271000000 μs/it ± 34000000

Summary:
  Unit growth: 18.1x, 16.1x, 16.0x
  Run time growth: 7.4x, 15.3x, 16.7x

Since there are various cases where the DOM parser exhibits slightly
different semantics, it is preferable to use the internal parser in
Scala.js too.