Fix number lexer to avoid conflict with object indexing

[itchyny]

This fixes #1526, also fixes #1651.
Currently jq does not correctly handle indexing like .e0. It tries to parse as numeric literal even if it does not accept it. I fixed the lexer to consider .e0 (and alike; .E1, .E-1) as object indexing, not invalid numeric literal.

 % echo '{"e0": 1, "E1": 2, "E": 3}' | jq '.e0, .E1, .E-1, .E+1'
jq: error: Invalid numeric literal at EOF at line 1, column 3 (while parsing '.e0') at <top-level>, line 1:
.e0, .E1, .E-1, .E+1
jq: error: Invalid numeric literal at EOF at line 1, column 3 (while parsing '.E1') at <top-level>, line 1:
.e0, .E1, .E-1, .E+1
jq: error: Invalid numeric literal at EOF at line 1, column 4 (while parsing '.E-1') at <top-level>, line 1:
.e0, .E1, .E-1, .E+1
jq: error: Invalid numeric literal at EOF at line 1, column 4 (while parsing '.E+1') at <top-level>, line 1:
.e0, .E1, .E-1, .E+1
jq: 4 compile errors

[coveralls]

Coverage increased (+0.4%) to 84.525% when pulling 9eec420 on itchyny:fix-number-lexer-indexing into 80052e5 on stedolan:master.

[wader]

Nice test

[itchyny]

Thank you.