`DfContext` HTML entity decoding is incomplete and introduces platform-specific inconsistencies

Problem

DfContext (which decodes HTML entities in defect @context) decodes incompletely and platform-dependently, so contexts can contain literal </" remnants and differ between operating systems.

Root cause

src/main/java/org/eolang/lints/DfContext.java:60-64:

this.context = defect.context()
    .replace("
", System.lineSeparator())
    .replace(""", "\"")
    .replace("&", "&");

1. & is replaced last, so double-encoded input like < becomes < (still encoded) instead of <.
2. The standard entities " and '/' are not handled (only "), so a context using " keeps the literal ".
3. 
 is replaced with System.lineSeparator() (\r\n on Windows, \n on Unix), while XML normalization always means LF — the resulting context differs across OSes and breaks string comparisons.

Minimal example

A context value <o> should decode to <o> but yields <o>; a value "x" keeps "x" verbatim.

Expected behavior

Decode entities in a single normalized pass:

• replace & first (or loop until stable),
• handle ", ', ', 
/
,
• use a fixed \n for line breaks, not System.lineSeparator().

Related

• src/test/java/org/eolang/lints/DfContextTest.java — only covers the happy path; the double-encoding and entity variants are untested.

Start with src/main/java/org/eolang/lints/DfContext.java:60-64 to trace the current entity decoding, then read src/test/java/org/eolang/lints/DfContextTest.java. Add coverage for double-encoded entities, quote and apostrophe variants, and normalized line breaks. Done means these cases decode consistently to the expected context on every operating system.