[core] Use babel-plugin-optimize-react

[eps1lon]

DONT_MERGE_BEFORE_https://github.com/facebook/create-react-app/pull/6219_OR_YOU_WILL_BE_FIRED

Closes #15238

[mui-pr-bot]

@material-ui/core: parsed: -0.57% 😍, gzip: +0.78%
@material-ui/lab: parsed: -0.43% 😍, gzip: +0.24%

Details of bundle changes.

Comparing: 14bbd57...89d372a

bundle	parsed diff	gzip diff	prev parsed	current parsed	prev gzip	current gzip
@material-ui/core	-0.57%	+0.78% 🔺	329,629	327,741	90,119	90,824
@material-ui/core/Paper	-0.01%	+0.09% 🔺	68,673	68,667	20,470	20,488
@material-ui/core/Paper.esm	+0.02% 🔺	+0.10% 🔺	62,058	62,070	19,209	19,229
@material-ui/core/Popper	-0.04%	-0.01%	29,185	29,173	10,434	10,433
@material-ui/core/Textarea	-0.49%	+0.30% 🔺	5,759	5,731	2,368	2,375
@material-ui/core/TrapFocus	-0.74%	+0.19% 🔺	3,806	3,778	1,608	1,611
@material-ui/core/styles/createMuiTheme	0.00%	0.00%	16,389	16,389	5,824	5,824
@material-ui/core/useMediaQuery	0.00%	0.00%	3,221	3,221	1,312	1,312
@material-ui/lab	-0.43%	+0.24% 🔺	152,680	152,031	46,546	46,656
@material-ui/styles	-0.02%	+0.13% 🔺	51,390	51,380	15,286	15,306
@material-ui/system	0.00%	0.00%	15,753	15,753	4,380	4,380
Button	-0.07%	+0.06% 🔺	79,426	79,367	24,280	24,294
Modal	-0.43%	+0.25% 🔺	14,963	14,898	5,218	5,231
Portal	0.00%	0.00%	3,579	3,579	1,568	1,568
Rating	-0.28%	+0.10% 🔺	70,735	70,537	22,079	22,100
Slider	-0.19%	+0.06% 🔺	75,071	74,930	23,272	23,285
colorManipulator	0.00%	0.00%	3,904	3,904	1,543	1,543
docs.landing	0.00%	0.00%	52,044	52,044	13,816	13,816
docs.main	0.00%	0.00%	590,571	590,571	188,751	188,751
packages/material-ui/build/umd/material-ui.production.min.js	-0.48%	+0.99% 🔺	299,715	298,269	86,127	86,980

Generated by 🚫 dangerJS against 89d372a

[eps1lon]

@material-ui/core: parsed: -0.53% heart_eyes, gzip: +0.77%

Expected result. Without benchmarks (parse time, property access perf) it's hard to tell if the longer load time over the wire is justified. Need a way to compare sizes using brotli but I'm missing usage numbers for that as well.

[eps1lon]

In it's current form not optimal. An in-house fork (#16860) is likely the better option.