Improve compilation performance

[Stranger6667]

1. Avoid string allocation during scope parsing & do not re-parse DEFAULT_ROOT_URL

Before, the string was cloned anyway - either DEFAULT_ROOT_URL or the one from the schema and then passed to Url::parse where it was cloned again.

Benches:
boolean_false compile false: -72% (212.48 ns -> 59.542 ns)
jsonschema big schema compile: -7.1% (3.4517 us -> 3.2085 us)
jsonschema small schema compile: -5.6% (3.3137 us -> 3.1238 us)

This behavior is reproducible over other benchmarks. It will have the most effect on the (re)compilation of small schemas + frequent use jsonschema::is_valid without reusing compiled schema.

2. $ref and boolean validators are compiled without calling push on a vector, i.e. the vector is created more efficiently

It gives ~9.6% speedup for boolean validators and really small speedup for $ref (probably need a better benchmark). There is a visible improvement in the perf output.

3. Avoid to_string call during format validators compilation

It gives ~11% improvement

4. Mark all compile calls with #[inline]

On validators that don't involve regex compilation, it gave from 3 to 25% improvement

5. Reusing DEFAULT_SCOPE in resolve_fragment

Not really about compilation, but it is 25-30% of improvement during reference resolving on various benches

6. Replacing FormatValidator with generated structs

Improves compilation for this keyword by 7-8% and makes validation slightly faster (probably due to removing a func call overhead). But for validate there is an extra match, which seems to be slightly faster than a function call overhead. I'll cleanup duplication later

[codecov]

Codecov Report

Merging #62 into master will decrease coverage by 0.71%.
The diff coverage is 85.18%.

@@            Coverage Diff             @@
##           master      #62      +/-   ##
==========================================
- Coverage   84.44%   83.72%   -0.72%     
==========================================
  Files          45       45              
  Lines        3143     3176      +33     
==========================================
+ Hits         2654     2659       +5     
- Misses        489      517      +28     

Impacted Files	Coverage Δ
src/keywords/additional_items.rs	95.16% <ø> (ø)
src/keywords/additional_properties.rs	86.29% <ø> (ø)
src/keywords/boolean.rs	81.48% <ø> (-18.52%)	⬇️
src/keywords/const_.rs	100.00% <ø> (ø)
src/keywords/contains.rs	96.55% <ø> (ø)
src/keywords/enum_.rs	95.83% <ø> (ø)
src/keywords/exclusive_maximum.rs	96.15% <ø> (ø)
src/keywords/exclusive_minimum.rs	96.15% <ø> (ø)
src/keywords/if_.rs	93.18% <ø> (ø)
src/keywords/items.rs	88.46% <ø> (ø)
... and 37 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2b57fa7...99f19aa. Read the comment docs.

[Stranger6667]

So, for some more or less realistic schemas, this PR gives 8-12% improvement in the compilation time and from 40% to 80% improvement for small single-keyword schemas