Optimize metalanguage parser (#92)

[jhnaldo]

This PR resolved issue #92.
Before applying this optimization, the evaluation result of Extractor for es2022 is as follows:

$ esmeta extract -extract:eval -silent
# extracting specification... (5,179 ms)
# - # of actual parsing: 3,240,105
# - # of using cached result: 1,889,181

However, after applying this optimization, the evaluation results becomes as follows:

$ esmeta extract -extract:eval -silent
# extracting specification... (4,288 ms)
# - # of actual parsing: 2,958,742
# - # of using cached result: 1,344,528