Add performance tests for 6502 implementations

[iflan]

There are currently 5 CPU implementations:

• jscpu6502.js: The original JS implementation of cpu6502.
• tscpu6502.ts: The initial Typescript conversion of the JS implementation using bind.
• tscpu6502v2.ts: A Typescript version that uses call instead of bind.
• tscpu6502v5.ts: A Typescript version that uses a giant switch statement and this instead of an array. (Only emulates the 6502.)
• tscpu6502v6.ts: A Typescript version using class variables from PR Use class fields instead of binding #40.

The performance tests added by this change use the following methodology:

• Tests are compiled and run using the production tool chain, including Webpack.
• Performance is measured by running the 6502 and 65C02 variations against the same ROMs as the unit tests.
• A configurable number of "warm-up" runs are performed before the actual measurement runs. This should cause the browser to optimize all of the methods it's going to optimize.
• A configurable number of measurement runs are performed.

Results include:

• Minimum time.
• Maximum time.
• Mean (average) time.
• Median time.
• Number of runs.

The CPU implementations live in test/perf/impl. Tests for the implementations live in test (though not all implementations have tests).

To run the tests using Node's http-server:

npm run build
http-server ./

then navigate to http://localhost:8080/test/perf/cpu_benchmark.html.

Note that there is plenty of room for improvement in both the benchmark implementation and the measurements.

Here are the results for an example run in Chrome 86 comparing the original JS version, the Typescript version using bind and the version from PR #40:

	cpu6502.js	cpu6502.js	cpu6502.ts	cpu6502.ts	cpu6502v6.ts	cpu6502v6.ts
	6502	65C02	6502	65C02	6502	65C02
min	1850.06	2227.38	1380.41	1613.08	1219.65	1547.14
max	2043.47	2303.67	1445.7	1697.81	1268.03	1624.13
mean	1881.92	2259.27	1403.78	1651.35	1237.17	1577.89
median	1874.9	2260.42	1397.07	1647.96	1233.75	1575.78
runs	50	50	50	50	50	50